ES2282136T3 - PROCEDURE AND APPARATUS TO PRODUCE A PRINT OF A BAYER IMAGE. - Google Patents

Abstract

A procedure for providing an image for printing at a predetermined bi-level point resolution corresponding to a predetermined continuous tone resolution, the method including the steps of: receiving a first set of data indicative of the image, the first set being data in planarized Bayer format comprising different color planes (45, 46, 47), at least one (46) of the color planes having a resolution different from that of the other plane or the other color planes (45, 47) ; Perform image reconstruction and resampling (64) of each color plane of the first data set, to create a second set of predetermined continuous tone resolution data that is greater than the resolution of each of the color planes of the first set of data; convert (67) the second data set into a third data set of the predetermined bi-level point resolution that is greater than the predetermined continuous tone resolution of the second data set, and make the third data set found available for a printer (69) at the default bi-level point resolution.

Description

Procedure and apparatus for producing a Bayer image printing.

Field of the Invention

The present invention relates to a procedure and to an apparatus for producing an impression of a Bayer image.

The invention has been developed mainly for a digital camera that includes an integral printer for provide a print on paper of an image captured by the camera, and it will be described in what follows with reference to that application. However, it will be appreciated that the invention is not limited to that particular field of use.

Summary of the invention

According to a first aspect of the invention, a method is provided for the provision of a  image for printing at dot resolution default bi-level, which corresponds to a default continuous tone resolution, including Procedure the stages of:

receive a first set of indicative data of the image, being the first data set in Bayer format by planes, comprising different planes (45, 46, 47) of colors, being at least one of the planes (46) of a color of one resolution different from that of the plane or planes (45, 47) of another color,

perform a reconstruction and resampling (64) of the image of each color plane of the first data set to create a second set of tone resolution data default continuous, that is greater than the resolution of each one of the color planes of the first data set;

convert (67) the second set of data into a third set of point resolution data default bi-level, which is greater than the default continuous tone resolution of the second set of data, and

make the third set of data be find available for a printer (69) at resolution Default bi-level points.

Preferably, the first resolution is equivalent to the default bi-level point resolution. In other embodiments, however, the first resolution is higher than the default bi-level point resolution.  In yet another embodiment, the first resolution is less than the default bi-level point resolution.

Also preferably, the first set of data is in red, green and blue RGB format, and the printer acts in response to a cyan, magenta and yellow CMY format, and the procedure includes the additional stage of converting the third data set from RGB format to CMY format.

In a preferred form, the procedure includes the stage of making the second set of data more intense. Alternatively, the procedure includes the preferred step of make the first set of data more intense.

Preferably, the first data set is obtained from a sensor device, and the procedure includes the stage of compensating the first set of data as soon as to the nonlinearities of the sensor device. More preferably, the offset stage includes converting the first data set from a plurality of samples of bits x to a plurality of bit samples y, where x> y. Even more preferred, x = 10 and y = 8.

Also preferably, the procedure includes the stage of arranging the first data set in plans by a red plane, a green plane and a blue plane.

In a preferred form, the procedure includes the additional stages of: determine for the first data set, m% of darker pixels and n% of pixels brighter;

adjust the first data set to match m% of darker pixels, and

adjust the first data set to match The n% of brightest pixels.

Preferably, the procedure includes the additional step of adjusting the first set of data to provide a predetermined white balance. More preferably, the method includes the additional step of adjusting the first set of data to provide a predetermined range expansion. Even more preferably, the color resolution of the first data set is increased while maintaining the same resolution.
space.

In a preferred form, the first set of data is selectively adjusted to provide the image with a default rotational orientation.

According to a second aspect of the invention, an apparatus is provided for the provision of a image, for printing at a predetermined resolution of points bi-level that corresponds to a resolution Default continuous tone, including the device:

input means for receiving a first set of data indicative of the image, the first set of data being in Bayer format by planes, comprising planes (45, 46, 47) of different colors being at least one of the planes (46) of color of a resolution different from that of the other plane (s) (45, 47) of
color,

sampling means for reconstruction and resampling (64) of image in the first data set, to create  a second set of data with a continuous tone resolution default, which is greater than the resolution of each of the color planes of the first data set;

processing means to convert (67) the second set of data in a third set of data with a default resolution of bi-level points which is greater than the default continuous tone resolution of the second data set, and

exit means to make the third data set available (69) for printing at Default resolution of bi-level points.

Preferably, the first resolution is matches the default resolution of points bi-level Alternatively, the first resolution is greater than the default resolution of points bi-level In another embodiment, however, the first resolution is less than the default resolution of bi-level points

Also preferably, the first set of data is in a red, green and blue (RGB) format, and the printer acts in response to a cyan, magenta and yellow (CMY) format, in which the processing means convert the third set of data from the RGB format to the CMY format.

Also preferably, the device does more Intense the second set of data. In another embodiment, without However, the apparatus intensifies the first set of data.

In a preferred form, the first data set is obtained from a sensor device, and the input means compensates for the first data set in terms of the nonlinearities of the sensor device. More preferably, compensation for nonlinearities includes converting the first set of data, from a plurality of samples of bits x, to a plurality of samples of bits y, where x> y. Even more preferably, x = 10, e
y = 8.

Preferably, the input means have in planes the first set of data according to a red plane, a plane Green and a blue plane.

More preferably, the input means:

determine, for the first set of data, the m% of darker pixels and n% of brighter pixels;

adjust the first data set to match m% of darker pixels, and

adjust the first data set to match The n% of brightest pixels.

In a preferred form, the input means adjust the first set of data to provide a balance of the default target. More preferably, the means of input adjust the first data set to provide a default range expansion. Even more preferably, the input media increase the color resolution of the first data set while maintaining the same resolution space.

Preferably, the input means adjust selectively the first set of data to provide the image with a default rotational orientation.

According to a third aspect of the Invention, a camera is provided that includes:

a CCD sensor device for the provision of a Bayer image;

a printer to selectively provide a printed image, and

an apparatus as described above, to receive the Bayer image and provide the printer with third set of data so that the image can be produced printed.

       \ newpage
    

Brief description of the drawings

Preferred embodiments of the invention go to be described now, by way of example only, with reference to the description and the following Figures.

Figure 1 shows a high image flow PCP level;

Figure 2 shows a diagram of the PCP by separated;

Figure 3 shows a block diagram of the PCP connected to the Printcam hardware;

Figure 4 shows a printhead Memjet 10 cm (4 inches);

Figure 5 shows the arrangement of segments on a 10 cm (4 inch) printhead;

Figure 6 shows the arrangement of nozzles in a pod, numbered in firing order;

Figure 7 shows the arrangement of nozzles in a pod, numbered by load order;

Figure 8 shows a chrome-pod;

Figure 9 shows a pod group;

Figure 10 shows a phase group;

Figure 11 shows the relationship between segments, trigger groups, phase groups, group-pods and chrome-pods;

Figure 12 shows pulse profiles of AEnable and BEnable during printing of an even and odd dot;

Figure 13 shows the orientation of formats printing based on the CFA image;

Figure 14 shows a block diagram of the image capture chain;

Figure 15 shows the pixel arrangement in a Bayer CFA 2G mosaic;

Figure 16 shows the RGB process of linearization;

Figure 17 shows the RGB process of planarization;

Figure 18 shows a block diagram of the image printing chain;

Figure 19 shows a color range of sample for a simple color plane;

Figure 20 shows the stages involved in white balance and range expansion;

Figure 21 shows a block diagram of a device capable of performing the white balance and the range expansion;

Figure 22 shows the various pixels of color plane in relation to a CFA resolution;

Figure 23 shows the effect of rotating the plane from green at 45 degrees;

Figure 24 shows the distance between rotated pixels for the green plane;

Figure 25 shows the movement process of mapping in CFA space not rotated with respect to CFA space turned;

Figure 26 shows a block diagram of the intensification process;

Figure 27 shows the process involved in high-pass filtering of a luminance pixel simple with a 3 x 3 function;

Figure 28 shows the transformation of RGB to CMY conversation;

Figure 29 shows the conversion of RGB to CMY by tri-linear interpolation;

Figure 30 shows the pixel replication of a simple pixel to a 5 x 5 block;

Figure 31 shows a block diagram of the halftone generation process;

Figure 32 shows the process of points of reformatting for the printer;

Figure 33 shows a block diagram of the image capture unit;

Figure 35 shows a block diagram of the image access unit;

Figure 36 shows a block diagram of the image histogram unit;

Figure 37 shows a block diagram of the printer interface;

Figure 38 shows the block diagram of the Memjet interface;

Figure 39 shows the generation of widths of AEnable and BEnable pulse;

Figure 40 shows a block diagram of point counting logic;

Figure 41 shows the interface of the unit print generator;

Figure 42 shows a block diagram of the print generator unit;

Figure 43 shows a block diagram of the test pattern access unit;

Figure 44 shows a block diagram of Intermediate Memory 5;

Figure 45 shows a block diagram of Intermediate Memory 4;

Figure 46 shows a block diagram of the UpInterpolate, Halftone and Reformat process;

Figure 47 shows how to perform the mapping from a vibrating cell to a vibrating cell accumulated

Figure 48 shows a block diagram of the Convert RGB to CMY process;

Figure 49 shows a block diagram of Intermediate Memory 2;

Figure 50 shows a basic spatial filter high pass that uses a 3 x 3 function;

Figure 51 shows a block diagram of the intensification unit;

Figure 52 shows the structure of the memory Intermediate 1;

Figure 53 shows a block diagram of the Resample and Create Luminance Channel process;

Figure 54 shows a block diagram of the Convolution Unit;

Figure 55 shows the pixel order generated from the receiver;

Figure 56 shows the movement in "x" or in "and" in the space turned and not turned;

Figure 57 shows the address of entries in the green sub-memory of the Buffer one;

Figure 58 shows the relationship between green entries based on rotation;

Figure 59 shows a 4x4 sampling of the channel of the green;

Figure 60 shows a 4x4 sampling, type 1, of the green;

Figure 61 shows a 4x4 sampling, type 2, of the green;

Figure 62 shows the two types of row addressing for green;

Figure 63 shows the addressing of entries in the red and blue sub-memories of buffer 1;

Figure 64 shows the first 16 samples read to calculate the first pixel;

       \ newpage
    

Figure 65 shows the worst case of 4 x 4 read overlay from the buffers of the blue and red;

Figure 66 shows a block diagram of the unit of rotation, white balance and range expansion, and

Figure 67 shows the active image area inside the generated coordinate space.

       \ vskip1.000000 \ baselineskip
    

1. Compendium of the PCP 1.1 High Level Functional Compendium

The Central Processor of the Printcam (PCP) has all the processing power for a Printcam, and it has specifically designed to be used in the camera system Digital Printcam. PCP 3 connects an image sensor 1 (for image capture), and a 2 Memjet printer for printing images. In terms of image processing, you can think that the PCP is the image translator from capture to printing, as shown in Figure 1:

?

        Image Sensor 1 is a CMOS image sensor, which captures a 1500 x 1000 RGB image. The Image Sensor constitutes the input device.

?

        The Print Head 2 is a 1600 dpi 10.16 cm (4 inches) Memjet printer long, able to print in three colors: cyan, magenta and yellow. The Print Head is the device for image output

?

        PCP 3 takes a picture from Image Sensor 1, process it, and send the final form of The image to Print Head 2, for printing. Since Image Sensor 1 captures in RGB, and Print Head 2 print in CMY, PCP 3 must translate from RGB color space up to CMY color space. PCP 3 contains all requirements for intermediate image processing, including White balance, color correction and range mapping frequency, image intensification, and generation of semitone. Additionally, PCP 3 controls the user interface and the entire printing process, providing support for a diversity of image formats PCP 3 also contains interfaces to allow export and import photos, complying with DPOF standard (Digital Print Order Format, Digital Format Print Orders).

       \ vskip1.000000 \ baselineskip
    

1.2 Internal High Level Compendium

PCP 3 is designed to be manufactured using a 0.25 micrometer CMOS process, with approximately 10 million transistors, almost half of which are memory Flash or static RAM. This leads to an estimated area of 16 mm2. The estimated manufacturing cost is 4 \ textdollar in 2001. The PCP 3 is a relatively straightforward design, and the design effort can be reduced with the use of techniques compilation of data paths, macro-cells and IP cores PCP 3 contains:

?

        One CPU / core 10 low speed microcontroller

?

        1.5 MBytes of Memory 11 multi-level flash (2 bits per cell)

?

        an Interface 98 of CMOS Image Sensor inside a Capture Unit Picture 12

?

        16 KBytes of memory flash 13 for program storage

?

        4 KBytes of RAM 14 for variable storage of programs.

PCP 3 has been planned to run to a clock speed of approximately 100 MHz to 3 V externally and 1.5 V internally to minimize power consumption. The actual operating frequency will be an integer multiple of the frequency Operation of the Print Head. CPU 10 is expected to be a simple microcontroller CPU, which runs to approximately 1 MHz. Both CPU 10 and sensor interface 12 CMOS, can be cores supplied by a supplier.

Figure 2 shows a block diagram of the PCP 3 separately.

PCP 3 is designed for use in systems from Printcam. Figure 3 shows a block diagram of PCP 3 connected to the rest of the Printcam hardware.

       \ newpage
    

2. Background of the printhead

PCP 3 has been specifically designed to connect to a print head 2 Memjet of 10 cm (4 inches).  Printhead 2 is used as a page printer at wide, producing a 10 cm (4 inch) printed image of width without having to be moved. On the contrary, paper 20 is print as it moves as it passes through printhead 2, as shown in Figure 4.

2.1 Print Head Composition of 10 cm (4 inches)

Each print head 2 of 10 cm (4 inches) consists of 8 segments, each segment being 12.7 mm (1/2 inch) in length. Each of the segments 21 prints cyan, magenta and yellow bi-level points on a different part of the page, to produce the final image. Segment positions are shown in the
Figure 5

Since printhead 2 prints dots at 1600 dpi, each point has a diameter of 22.5 µm, and are separated by 15,875 µm. Thus, each 12.7 mm segment (half an inch) prints 800 dots, corresponding to the 8 segments to positions:

TABLE 1 Final Image Points Directed by Each Segment

one

Although each segment 21 produces 800 points of the final image, each point is represented by a combination of cyan, magenta and yellow bi-level ink. Since The printing is bi-level, the input image will be Vibrated or diffuse error for better results.

Each segment 21 then contains 2,400 nozzles: 800 of each color cyan, magenta and yellow. A head Print 2 of 10 cm (four inches) contains 8 segments 21 with a total of 19,200 nozzles.

2.1.1 Grouping of Nozzles Within a Segment

The nozzles 22 inside a segment 21 simple, they are grouped for reasons of physical stability as well as well as reduction of power consumption during Print. In terms of physical stability, a total of 10 nozzles share the same ink tank. In terms of power consumption, the clusters are made so that allow a low speed and a high pressure mode speed.

Printhead 2 supports two speeds printing to allow different ones to be carried out speed / power ratios in different configurations of products.

In low speed printing mode, 96 nozzles 22 are fired simultaneously from each head of 10 cm (4 inch) print 2. The fired nozzles must be maximally distant, so that 12 nozzles are fired 22 from each segment. To shoot all 19,200 nozzles, it 200 different sets of 96 nozzles must be fired.

In high speed printing mode, it they shoot 192 nozzles 22 simultaneously from each head of 10 cm (4 inch) print 2. The nozzles 22 fired must be maximally distant, so that 24 nozzles are fired from each segment. To fire the entire 19,200 nozzles, 100 different sets of 192 nozzles must be fired.

Power consumption in low mode Speed is half that in high speed mode. Observe, however, that the energy consumed to print a line, and by both a page, is the same in both cases.

In a scenario such as a powered Printcam with battery, the power consumption requirements determine the use of low speed printing.

2.1.1.1 10 Nozzles Form a Pod

A simple pod 23 consists of 10 nozzles 22 that They share a common 5 ink tank. 5 nozzles 22 are in one row, and 5 are in another. Each nozzle 22 produces points of approximately 22.5 µm in diameter, separated on a 15,875 µm grill. Figure 6 shows the layout of a simple pod, with the nozzles 22 numbered according to the order in which they must be shot.

Although nozzles 22 are fired by this order, the ratio of nozzles 22 and the physical placement of points On the printed page, it is different. Nozzles 22 in a row they represent the even points of a page line, and the nozzles in the other row represent the odd points of the line adjacent to the page. Figure 7 shows the same pod 23 with the nozzles 22 numbered according to the order in which the They must be loaded.

The inner nozzles 22 to a pod 23, are therefore logically separated by the width of 1 point. The exact distance between the nozzles 22 will depend on the properties of the Memjet trigger mechanism. Printhead 2 is designed with staggered nozzles designed to match the paper flow 20.

2.1.1.2 3 Pods Form a Chroma-Pod

A pod 23 of each color (cyan, magenta and yellow), are grouped in a chroma-pod 24. A Chroma-Pod 24 represents components of different color of the same horizontal set of 10 points, on lines different. The exact distance between different color pods 23 It depends on the operating parameters of the Memjet, and may vary from one Memjet design to another. The distance is considered to be a constant amount of knit widths, and should be taken in account therefore when printing is done: the dots printed by cyan nozzles will be for different lines that those printed by the magenta or yellow nozzles. The algorithm of  printing should allow a variable distance of up to approximately 8 knit widths between colors (see Table 3 for more details). Figure 8 illustrates a chroma-pod 24 simple.

2.1.1.3 5 Chroma-Pods Form a Group-Pod

5 chroma-pods 24 are organized in a single group-pod 25. Since each Chroma-pod contains 30 nozzles 22, each pod-group contains 150 nozzles 22: 50 nozzles cyan, 50 magenta nozzles and 50 yellow nozzles. The disposition It is shown in Figure 9, with the chroma-pods numbered 0-4. Note that the distance between adjacent chroma-pods has been exaggerated by reasons for clarity

2.1.1.4 2 Group-Pods Form a Group-Phase

2 group-pods 25 are organized in a single group-phase 26. The group-phase 26 is named in that way because the groups of nozzles 23 inside a group-phase are fired simultaneously during a given trigger phase (this is explained in more detail in what follow). The formation of a group-phase from 2 group-pods 25, is done entirely for the purpose of low speed and high speed printing through 2 lines PodgroupEnable.

During low speed printing, it will has only one of the two PodgroupEnable lines in one pulse given shot, so that only one group-pod of both shoots the nozzles. During high printing speed, both PodgroupEnable lines are arranged, so that both group-pods shoot the nozzles. In Consequently, low speed printing is twice as long than high speed printing, since high printing Speed fires twice as many nozzles at once.

Figure 10 illustrates the composition of the group-phase The distance between Adjacent group-pods have been exaggerated for reasons of clarity

2.1.1.5 2 Group-Phases form a Shooting-Group

Two group-phases (PhasegroupA and PhasegroupB) are organized into a single firing group 27, with 4 firing groups in each segment. The firing groups 27 are so named because they all fire the same nozzles 27 simultaneously. Two enable lines, AEnable and BEnable, allow the tripping of the PhasegroupA nozzles and the PhasegroupB nozzles independently as different firing phases. The arrangement is shown in Figure 11. The distance between adjacent clusters has been exaggerated for reasons of
clarity.

2.1.1.6 Nozzle Grouping Summary

Table 2 is a summary of the groupings of nozzles on a printhead.

       \ vskip1.000000 \ baselineskip
    

TABLE 2 Nozzle clusters for a head of 10 cm (4 inch) simple print

2

2.2 Load and Print Cycles

A simple print head 2 of 10 cm (4 inches) contains a total of 19,200 nozzles 22. A Cycle of Printing includes shooting up to all of these nozzles, depending on the information to be printed. A Load Cycle includes loading the printhead with the information to be printed during the subsequent Cycle of Print.

Each nozzle 22 has a NozzleEnable bit associated, which determines whether the nozzle is to be fired or not during the Print Cycle. The NozzleEnable bits (one per nozzle) are loaded by means of a set of records of displacement.

Logically, there are 3 records of displacement per segment (one per color), each with a length of 800. As the bits move through the register of displacement for a given color, are directed until lower and upper nozzles on alternate pulses. Internally, each shift register with a depth of 800, It comprises two displacement registers of 400 depth: one for the upper nozzles and one for the lower nozzles. Alternate bits are shifted to internal registers. In as far as the external interface is concerned, however, there is a only 800-meter deep offset register.

Once all records of displacement have been fully charged (800 pulses of load), all bits are transferred in parallel to the bits of NozzleEnable appropriate. This amounts to a parallel transfer. Simple 19,200 bits. Once the transfer has taken place, The Print Cycle can start. The Print Cycle and the Load Cycle can occur simultaneously while the parallel loading of all NozzleEnable bits at the end of Print Cycle

2.2.1 Load Cycle

The Load Cycle is related to the load of print head offset records with the following NozzleEnable bits of the Print Cycle.

Each segment 21 has 3 related entries directly with the cyan displacement records, the Magenta and yellow. These entries are called CDatain, MDatain and YDatain. Since there are 8 segments, there are a total of 24 10 cm color input lines per printhead (4 inches). A simple pulse along the SRClock line (shared between all 8 segments), transfer the 24 bits to the appropriate displacement records. Alternate pulses transfer bits to the lower and upper nozzles, respectively. Market Stall that there are 19,200 nozzles, a total of 800 pulses is required For transfer. Once the entirety has been transferred of 19,200 pulses, a simple pulse along the PTransfer line shared, causes parallel data transfer from shift registers to the appropriate NozzleEnable bits.

Parallel transfer through a pulse on PTransfer, it must take place after the Print cycle Otherwise, the NozzleEnable bits for the line being printed will be incorrect.

Since the 8 segments 21 are loaded with a Simple SRClock pulse, any printing procedure must produce the data in the correct sequence for the head of Print. As an example, the first SRClock pulse will transfer the CMY bits for the next point 0, 800, 1600, 2400, 3200, 4000, 4800 and 5600 of the Print Cycle. The second SRClock pulse will transfer the CMY bits for the next point 1, 801, 1601, 2401,  3201, 4001, 4801 and 5601 of the Print Cycle. After 800 pulses SRClock, the PTransfer pulse can be provided.

It is important to note that odd CMY outputs and even, although printed during the same Print Cycle, no they appear on the same physical output line. Physical separation of odd and even nozzles inside the printhead, as well as the separation between nozzles of different colors, ensures that will produce points on different lines of the page. The relative difference must be taken into account when loading printhead data The real difference in lines depends on the characteristics of the inkjet mechanism that be used on the printhead. The differences may be defined by variables D1 and D2, where D1 is the distance between nozzles of different colors, and D_ {2} is the distance between nozzles of the same color. Table 3 shows the points transferred to segment n of a printhead over The first 4 pulses.

TABLE 3 Order of Points Transferred to a Head of 10 cm (4 inch) print

4

And so on for the entire 800 pulses

The data can be registered in the head of printing at a maximum speed of 20 MHz, which will be loaded all data for the next line at 40 \ mus.

2.2.2 Print Cycle

A print head 2 of 10 cm (4 inches) contains 19,200 nozzles 22. Shoot them all at once could consume too much power and be problematic in terms of ink refill and nozzle interference. In consecuense, two shooting modes are defined: a low print mode Speed and high speed printing mode:

?

        In print mode Low speed, there are 200 phases, firing each phase 96 nozzles This equates to 12 nozzles per segment, or 3 per group. Shooting

?

        In print mode High speed, there are 100 phases, firing each phase 192 nozzles This equates to 24 nozzles per segment, or 6 groups of Shooting.

The nozzles that are going to be fired during A given pulse is determined by:

?

        3 bit ChromapodSelect (select 1 of 5 chroma-pods 24 from a trigger group 27)

?

        4-bit NozzleSelect (select 1 of 10 nozzles 22 from a pod 23)

?

        2 line bits PodgroupEnable (select 0, 1 or 2 group-pods 25 for shooting).

When one of the lines is established PodgroupEnable, only the 4 specified nozzles of the Podgroup  will fire as determined by ChromapodSelect and NozzleSelect When both PodgroupEnable lines are determined, Both groups can fire their nozzles. For low mode speed, two firing pulses are required, with PodgroupEnable = 10 and 01, respectively. For high speed mode, only a trigger pulse is required, with PodgroupEnable = 11.

The duration of the trigger pulse is given by AEnable and BEnable lines, which trigger the nozzles PhasegroupA and PhasegroupB of all trigger groups, respectively. The typical duration of a trigger pulse is 1.3-1.8, respectively. The duration of a pulse depends on ink viscosity (dependent on temperature and characteristics of the ink), and the amount of power available with respect to the printhead. See Section 2.3 for details about reports from the printhead to temperature change compensation effects.

The AEnable and BEnable lines are lines separated so that the trigger pulses can overlap. Thus, the 200 phases of a Low Print Cycle Speed consist of 100 phases A and 100 phases B, which provide effectively 100 sets of Phase A and Phase B. Likewise mode, the 100 phases of a high speed printing cycle They consist of 50 phases A and 50 phases B, which provide effective 50 phases of phase A and phase B.

Figure 12 shows the AEnable and BEnable lines during a typical Print Cycle. In a high speed print, there are 50 cycles of 2 \ mus, while in a low speed print there are 100 cycles of
2 \ mus.

For high speed printing mode, the firing order is:

?

        ChromapodSelect 0, NozzleSelect 0, PodgroupEnable 11 (Phases A and B)

?

        ChromapodSelect 1, NozzleSelect 0, PodgroupEnable 11 (Phases A and B)

?

        ChromapodSelect 2, NozzleSelect 0, PodgroupEnable 11 (Phases A and B)

?

        ChromapodSelect 3, NozzleSelect 0, PodgroupEnable 11 (Phases A and B)

?

        ChromapodSelect 4, NozzleSelect 0 PodgroupEnable 11 (Phases A and B)

?

        ChromapodSelect 0, NozzleSelect 1, PodgroupEnable 11 (Phases A and B)

?

        ......

?

        ChromapodSelect 3, NozzleSelect 9, PodgroupEnable 11 (Phases A and B)

?

        ChromapodSelect 4, NozzleSelect 9, PodgroupEnable 11 (Phases A and B)

For low speed printing mode, the firing order is similar. For each phase of the high mode speed, in which PodgroupEnable was 11, two phases are replaced of PodgroupEnable = 01 and 10, as follows:

?

        ChromapodSelect 0, NozzleSelect 0, PodgroupEnable 01 (Phases A and B)

?

        ChromapodSelect 0, NozzleSelect 0, PodgroupEnable 10 (Phases A and B)

?

        ChromapodSelect 1, NozzleSelect 0, PodgroupEnable 01 (Phases A and B)

?

        ChromapodSelect 1, NozzleSelect 0, PodgroupEnable 10 (Phases A and B)

?

        ......

?

        ChromapodSelect 3, NozzleSelect 9, PodgroupEnable 01 (Phases A and B)

?

        ChromapodSelect 3, NozzleSelect 9, PodgroupEnable 10 (Phases A and B)

?

        ChromapodSelect 4, NozzleSelect 9, PodgroupEnable 01 (Phases A and B)

?

        ChromapodSelect 4, NozzleSelect 9, PodgroupEnable 10 (Phases A and B)

When a nozzle 22 is fired, it takes about 100 \ mus to be filled. The nozzle 22 cannot be fired before this time has elapsed filled in. This limits the fastest print speed to 100 \ mus per line. In high speed printing mode, the time to print a line is 100 \ mus, so that the time between firing a nozzle from one line to the next is matched at the time of refilling, which makes the mode of High speed printing be acceptable. The printing mode of Low speed is slower than this, so it is also acceptable.

The shot of a nozzle 22 also causes acoustic disturbances for a limited time indoors of a common ink reservoir of pod 23 of the nozzle. The disturbances may interfere with the firing of another nozzle inside the same pod 23. Consequently, the shot of nozzles within a pod should be separated from each other as much as possible. Therefore, three nozzles of a chroma-pod 24 (one nozzle 22 per color) and a then move to the next chroma-pod 24 within the pod-group 25.

?

        In print mode low speed, pod-groups are fired by separated. Thus, the 5 chroma-pods inside both group-pods must be shot all before the first chroma-pod is triggered again, totaling 10 x 2 \ mus cycles. Consequently, each pod 23 fires a once for every 20 \ mus.

?

        In print mode high-speed, group-pods 25 soar jointly. That way, the 5 chroma-pods 24 interiors to a simple group-pod must be shot all before the first chroma-pod is triggered again, totaling cycles of 5 x 2 \ mus. Consequently, each pod 23 trips once every 10 \ mus.

Since the ink channel is 300 µm long and the speed of sound in the ink is around 1500 m / s, The resonance frequency of the ink channel is 2.5 MHz, so that the low speed mode allows 50 resonance cycles to Dampen the acoustic pulse, and the high speed mode allows 25 resonance cycles Thus, any acoustic interference is minimum in both cases.

2.2.3 Sample Time

As an example, consider printing time of a 10 x 15 cm (4 '' x 6 '') photo in 2 seconds, as required by the Printcam. In order to print a photo in 2 seconds, the 10 cm (4 inch) printhead should print 9,600 lines (6 x 1,600). Rounding to 10,000 lines in 2 seconds produces a line time of 200 \ mus. A cycle of Simple printing and a single Load Cycle must end both within this time Additionally, the physical process external to the print head should move the paper a quantity appropriate.

From the point of view of printing, the mode Low speed printing allows a print head 10 cm (4 inches) print a complete line at 200 \ mus. In Low speed printing mode, 96 nozzles 22 are fired per trigger pulse, thereby enabling line printing Complete within the specified time.

The 800 SRClock pulses for the head of print 2 (each clock pulse transfers 24 bits), must have also place within the 200 \ mus line time. The length of a pulse SRClock cannot exceed 200 \ mus / 800 = 250 ns, what which indicates that the printhead must be controlled at 4 MHz. Additionally, the average time to calculate each bit value (for each of the 19,200 nozzles) must not exceed 200 \ mus / 19.200 = 10 ns. This requires a pulse generator that runs at one of the following speeds:

?

        100 MHz generating 1 bit (point) per cycle

?

         50 MHz generating 2 bits (points) per cycle

?

         25 MHz generating 4 bits (points) per cycle.

       \ vskip1.000000 \ baselineskip
    

2.3 Reports from the Print Head

Printhead 2 produces several lines of information (accumulated from the 8 segments). Lines information are used to adjust the pulse time of Shooting. Although each segment 21 produces the same information, the reports from all segments share the same tri-state bus lines. In consecuense, only a segment 21 can provide information in each moment.

A pulse through the composite SenseSegSelect line with data on Cyan, enable detection lines for that segment. The information detection lines will come from segment selected until the next SenseSegSelect pulse. The Information detection lines are as follows:

?

        Tsense informs the controller how hot the print head is. This allows the controller to adjust the pulse trigger time, since the temperature affects the viscosity of the ink.

?

        Vsense informs the controller on how much voltage is available for the actuator This allows the controller to compensate for a battery of flat response or a high voltage source by adjusting the width Pulse

?

        Rsense informs the heater resistivity controller (ohms per frame) Actuator This allows the controller to adjust the widths of pulse to maintain a constant energy regardless of the heater resistivity.

?

        Wsense informs the width controller of the critical part of the heater, the which may vary up to ± 5% due to variations Lithographic and chemical engraving. This allows the controller adjust the pulse width appropriately.

2.4 Special cycles 2.4.1 Preheat Cycle

The printing process has a strong tendency to remain at equilibrium temperature. To ensure  that the first section of the printed photograph has a size of uniform point, the equilibrium temperature must reach, with before printing, to all points. This is achieved by means of a preheating cycle.

The Preheat cycle includes a Load Cycle for all nozzles with 1s (i.e., disposing all nozzles for firing), and a number of short firing pulses for each nozzle. The pulse duration must be insufficient to trigger the drops, but sufficient to heat the ink. In total, approximately 200 pulses are required for each nozzle, performing cycles using the same sequence as a Print Cycle
standard.

The information during the mode of Preheating, is provided by Tsense, and continues until that an equilibrium temperature is reached (around 30ºC per above the environment). The duration of the Preheat mode is around 50 milliseconds, and it depends on the composition of the ink.

Preheating is done before each print job This does not affect the performance of the printer,  since it is being carried out while the data to the printer.

2.4.2 Cleaning Cycle

In order to reduce the chances of the nozzles are clogged, a cycle of cleaning before each print job. Each nozzle it fires a number of times towards an absorbent sponge.

The cleaning cycle includes a single cycle of Load for all nozzles with 1s (that is, arranging all the nozzles for firing), and a number of firing pulses for each nozzle The nozzles are cleaned by means of it nozzle firing sequence as in a Print Cycle standard. The number of times each nozzle fires depends on of the ink composition and the time the printer has idle state, so as with preheating, the cleaning cycle has no effect on the behavior of the printer.

       \ vskip1.000000 \ baselineskip
    

       \ vskip1.000000 \ baselineskip
    

       \ vskip1.000000 \ baselineskip
    

(Table goes to page next)

       \ newpage
    

2.5 Printhead interface summary

A print head 2 of 10 cm (4 inches) It has the following connections:

       \ vskip1.000000 \ baselineskip
    

TABLE 4 10cm Print Head Connections (four inches)

5

       \ newpage
    

Inside the printhead of 10 cm (4 inches), each segment has the connections that follow with link adapters:

       \ vskip1.000000 \ baselineskip
    

TABLE 5 Internal Segment Connections Head 10 cm (four inch) print

6

       \ vskip1.000000 \ baselineskip
    

3. Image processing chains

The previous sections are related only with the compendium of PCP functionality that is level higher than that of CFA image mapping with respect to a diversity of print output formats. In fact, there is a number of stages involved in taking an image from the sensor image, and in the production of a high output print quality. We have divided the high level process into two image processing chains, each with a number of stages:

?

        Capture Chain Image

?

        Chain of Print.

The Image Capture Chain refers to the image capture from the Image Sensor, and to your storage locally inside the Printcam. Chain Print is related to taking the stored image for printing it. These two chains map over the Basic Printcam functionality as follows:

?

        Take & Print = Chain Image Capture followed by the Print Chain

?

        Reprint = Chain of Print.

For example, a user can print a Thumbnail image (Take & Print), and if you are happy with the results, print several standard copies (Reprint).

The chapter describes an implementation of the independent image processing chain that meets the Printcam quality requirements. In this phase, we are not considering exactly how the processing is done in hardware terms but, on the contrary, what should be done. These functions must be mapped onto several units within the PCP

Regardless of the PCP implementation, There are a number of limitations:

?

        The input image is an RGB image contone based on CFA.

?

        The output image is for a Memjet printhead (bi-level points at 1,600 dpi) in a CMY color space, and always has the same output width (10 cm wide (4 inches)).

3.0.1 Supported Printing Formats

PCP 3 supports a variety of formats of print output, as shown in Table 6. In all cases, the width of the image is 10 cm (4 inches) (comparable to the width of the printhead). Only the print length

TABLE 6 Image Formats Supported

7

The image sensor does not provide information of the orientation. All input images are captured at the same resolution (1,500 x 1,000), and may need to be rotated 90 degrees before being printed. Figure 13 illustrates the mapping between the captured CFA image and the various formats of printing supported. Note that although the image is shown rotated 90 degrees counterclockwise, the image It can be turned for or against the hands of the clock.

3.1 Image capture chain

The Image Capture Chain is responsible for taking an image from the Image Sensor, and storing it locally inside the Printcam. The Capture Chain of Image includes a number of processes that only need to be carried out during image capture. The chain of Image Capture has been illustrated in Figure 14, with sections consequents detailing the sub-components.

3.1.1 Image Sensor 1

The input image comes from a sensor image 1. Although a variety of sensors, we have only considered the color filter set (CFA) Bayer. The CFA Bayer has a number of attributes that are here define.

The image captured by the 1 CMOS sensor (a through a pickup lens), it’s supposed to have been sufficiently filtered in order to remove any particles of overlap. The sensor itself has a ratio of 3: 2 aspect, with a resolution of 1,500 x 1,000 samples. The Pixel's most likely arrangement is the color filter set (CFA) Bayer, with each 2 x 2 pixel block arranged in a 2G mosaic as shown in Figure 15.

Each sample contains R, G or B (corresponding to red, green or blue, respectively), is 10 bits Note that each pixel of the mosaic contains information about only one of the R, G or B. Estimates on the Lack of color information must be done before the image It can be printed.

The CFA is considered to make a deletion adequate fixed configuration noise (FPN).

3.1.2 Linearize RGB 40

?

        The image sensor 40 It is unlikely to have a completely linear response. For the Therefore, 10-bit RGB samples from the CFA must be considered as nonlinear. These nonlinear samples are translated into 8-bit linear samples by means of tables of search (one table per color).

Pixels from CFA lines 0, 2, 4 etc, are indexed in the R and G tables, while the pixels from lines CFA 1, 3, 5, etc., are indexed in the tables of G and B. This is completely independent of the orientation of the camera. The process is shown in Figure 16. The total amount of memory required for each search table is 2 10 x 8 bits The 3 search tables 45 therefore require a total 3 Kbytes (3 x 2 10 bytes).

3.1.3 Planarize RGB 41

The pixels obtained from the CFA have their color planes interspersed due to the nature of the mosaic Bayer of pixels. By this we mean that on lines horizontal uniform, a red pixel is followed by a green pixel and then for another red pixel (the different color planes are interspersed with each other). In some systems of image processing, an interleaved format is highly Useful. However, in the Printcam processing system, the Algorithms are more effective if you work on planar RGB.

A planarized image consists of one that has been separated into its integral colors. In the case of the CFA RGB image, there are 3 separate images: an image that contains only the red pixels, an image that contains only the blue pixels, and an image that contains only the green pixels. Note that each plane represents only the pixels of that color that were actually sampled. No resampling is done during the planning process. As a result, the R, G and B planes are planes not registered with each other, and the G plane is twice as large as any of the R or B planes. The process is shown in the
Figure 17

The actual process is very simple; depending on color of the read pixels, the output pixels are sent to the next position in the image of the appropriate color plane (for therefore, with the same orientation as the CFA).

The red 45 and blue 47 planar images are exactly a quarter of the size of the original CFA image. These are exactly half the resolution in each dimension. The red and blue images are therefore of
750 x 500 pixels each, with a red image implicitly offset from the blue image by a pixel in the CFA space (1,500 x 1,000) in both x and y dimensions.

Although the green 45 planar image is half in size than the original CFA image, it is not set directly like the red or blue planes. The reason is due to the extension in Green chess board shape. In a line, green is each odd pixel, and on the next line, green is each pixel pair. These alternating lines of the green plane represent pixels Odd and even within the CFA image. That way, the image Green planar is 750 x 1,000 pixels. This one has ramifications for the resampling process (see "Resampling at 64" more ahead).

3.1.4 Stored Image 42

Each color plane of the linearized RGB image It is written in memory for temporary storage. The memory should be Flash 11 so that the image is maintained after the power has been interrupted.

The total amount of memory required for the Planarized linear RGB image, is 1,500,000 bytes (approximately 1.5 MB), organized as follows:

?

        R: 750 x 500 = 375,000 bytes

?

        B: 750 x 500 = 375,000 bytes

?

        G: 750 x 1,000 = 750,000 bytes

3.2 Print Chain

The Printing Chain refers to taking a image already existing from memory 42, and print it on a 2 Memjet printer. An image is typically printed as soon as as it has been captured, although it can be reprinted (that is, without recapture).

There are a number of stages that are required in the image processing chain in order to produce High quality prints from captured CFA images. The Figure 18 illustrates the Print Chain. The chain is divided into 3 work solutions The first is the 50 capture space of original image (the same space as the CFA), the second is the intermediate resolution 51 (1,280 pixel tone lines continuous), and the final resolution is printer resolution 52, with lines of 6,400 bi-level points.

3.2.1 Input Image

The input image is an RGB 42 image linearized stored in planar form, as stored by the Capture Chain that has been described in Section 3.1.4.

3.2.2 Accumulation Statistics 60

A number of statistics related to the full image need to be accumulated before it can perform processes such as white balance and range expansion These statistics need only be accumulated once for all prints of an image 42 captured particular, and can be accumulated separately planar images of red, green and blue.

3.2.2.1 Histogram Construction

The first stage is to build a histogram for each 8-bit value of the color plane. Each picture  1,500 x 1,000 CFA contains a total of:

?

        375,000 pixels red (minimum required 19-bit counter)

?

        375,000 blue pixels (minimum required 19-bit counter)

?

        750,000 pixels green (minimum required 20 bit counter).

Therefore, a simple table of 256 x 20 bits to maintain the histogram.

The histogram construction procedure It is direct, as illustrated by the following pseudo-code:

8

       \ vskip1.000000 \ baselineskip
    

3.2.2.2 Determine High and Low Thresholds

Once the histogram has been constructed for the color plane, it can be used to determine a high threshold and low threshold. These thresholds can be used to automate afterwards white balance and range expansion during the printing process.

Basing the thresholds on the number of pixels in the histogram, we consider that the n% of darker pixels are expandable and therefore equal. In the same way, we consider that the n% of brightest pixels are expandable and therefore same. The exact value of "n" is expected to be around 5%, but will depend on the response characteristics of CFA

The process of determining n% of values Darker is direct. This includes a stepping through the histogram of the color plane from counting to 0 ascending (that is, 0, 1, 2, 3, etc.) until the total n% is reached or we we have displaced more than a set amount from 0. The most high of these values is considered to be the low threshold of the plane color. Although there is some difference between these values more dark, the difference can be considered disposable to range expansion and color balancing effects.

The process of determining n% of values Brighter, it's similar. This one includes a stepping through of the color plane histogram from 255 count descending (i.e. 255, 254, 253, etc.) until the n% total or until we have moved more than one amount established since 255. The lowest of these values is considered the high threshold of the color plane. Although there is some difference Among these brightest values, the difference can be discarded for the purposes of range expansion and balancing of color.

The reason for the arrest after a distance established from 0 or 255, consists of compensating Two types of images:

?

        where the dynamic range original be low, or

?

        where there is no white or Black in the picture.

In these two cases, we do not want to consider that the entire n% of upper and lower values be disposable since we have a low range to start from. We can safely establish that thresholds high 73 and low 72 are outside the range of pixel values really sampled The exact distance will depend on the CFA, but it will consist in two constants.

A sample color range for a plane of color has been shown in Figure 19. Note that although it is possible the full range of 0-255 for pixels of the image color plane, this particular image has a more range small. Note also that the same n% of rank 70, 71 of histogram is represented by a larger range at the end 70 low than at the end 71 high. This is because the histogram must contain more pixels with higher values closer to each other That if compared to the low end.

The high 73 and low 72 thresholds must be determined for each color plane individually. This information may be used to calculate the range scale and deviation factors to be used in the process subsequent white balance and range expansion.

The following pseudo code illustrates the process of determining either thresholds (to find the low threshold, StartPosition = 255, and Delta = 1. To find the high threshold, StartPosition = 0 and Delta = -1). The pseudo-code assumes that Threshold is a value 8-bit that cyclically restarts during the addition

9

       \ vskip1.000000 \ baselineskip
    

3.2.3 Rotate Image 61

Image rotation 61 is a stage optional in both Capture and Print processes and Reprint.

Different print formats require that the image is rotated either 0 or 90 degrees in relation to the CAF orientation, as shown in Figure 13. The amount of Rotation depends on the currently selected print format. Although the direction of rotation is not important (it may be in favor clockwise or counterclockwise that the new orientation is only intended to facilitate the print head width), the direction of rotation it will affect the relative registration of the 3 color planes. Table 7 summarizes the rotation required for each print format from of the original CFA orientation.

TABLE 7 Rotations from the CFA orientation for Formats Print

10

       \ vskip1.000000 \ baselineskip
    

Since we are rotating only 90 degrees, No information is lost during the rotation process. For a rotation of 0, the image can be read in rows, and for a rotation of 90, the image can be read column by column. Registration of 3 color planes must take into account the address of rotation

3.2.4 White Balance and Range Expansion 63

A photograph is rarely taken in ideal lighting conditions. Even the notion of "perfect lighting conditions" is loaded with subjectivity, both in terms of the photographer and the object. Without However, in all cases, the object of the photograph is illuminated with light, either from a light source (such as the sun or a interior lighting), or with its own light (such as a neon sign).

In most of the conditions of lighting, it may seem to the photographer as light "white", which is usually far from being white light. Interior lighting, for example, typically has a tone yellow, and this yellow shade will lead to a photograph wrong. For most people, the yellow tone on The final wrong picture is wrong. Although it can complement the vision conditions at the time the photography, does not match the perceived color of the object. It turns out by both crucial to perform a white balance on a photograph before printing.

In the same way, an image can be perceived as of a higher quality when the dynamic range of the colors expands to match the full range in Each color plane. This is particularly useful to perform before that an image be resampled for a higher resolution. Yes dynamic range is higher, values can be used intermediate interpolated pixel positions, avoiding a staggered or fractional image. The range expansion is designed to provide the full 256 value range to those values really sampled. In the best case, the value lowest is mapped to 0, and the highest value is mapped to 255. All intermediate values are mapped to values proportionally intermediate between 0 and 255.

Mathematically, the operation performed consists of in a translation of Low Threshold 72 to 0, followed by a scaled up The formula is shown below:

       \ vskip1.000000 \ baselineskip
    

eleven

RangeScaleFactor must be limited to one maximum value to reduce the risk of range expansion too  far. For details on the calculation of LowThreshold 72, see Section 3.2.2 "Accumulated Statistics". This values (LowThreshold and RangeScaleFactor) will be different for each plane of color, and only need to be calculated once per image.

Both tasks can be undertaken simultaneously, as shown in Figure 20.

Since this stage includes a scalar process, we can leave some fractional component in the mapped value, for example the value 12 can be mapped up to 5.25. Instead of discarding the fractional component, we turn to a 10-bit result (8 bits of integer, 2 of fraction) in the next phase of the image processing chain. We cannot provide a memory to store the entire image at more than 8 bits, but we can make good use of the highest resolution in the resampling phase. Therefore, the input image is 8 bits, and the output image has 10 bits per color component. The logical process is shown in the
Figure 21

It is important to have a lower limit of 0 during subtraction so that all values below LowThreshold 72 will be mapped up to 0. Similarly, the multiplication must have a ceiling of 255 for the entire portion of the result, so that values higher than HighThreshold 73 will be mapped up to 255.

3.2.5 Resampling 64

The CFA provides only one component of Simple color per pixel coordinate (x, y). To produce the final printed image, we need to have the other values of color component in each pixel. Finally, we need the cyan, magenta and yellow components in each pixel, but to get to cyan, magenta and yellow we need red, the Green and blue With our one-color-per-pixel, we have the red component for a particular position, but We need to estimate blue and green. Or we can have the green and We need to estimate red and blue.

Even if we had the components of full red, green and blue color for each pixel resolution CFA, the CFA resolution image is not the final resolution of exit. Additionally, although the output format varies, the Physical width of the printed image is constant (10 cm (4 inches) at 1,600 dpi). The constant width of the printhead is per 6,400 points.

There are two extreme cases that must be to consider:

-

                    Interpolate at CFA resolution (minimum interpolation), and then perform intensification, color conversion. Finally, extrapolation to print resolution. This has the advantage of a constant intensification function and a Low resolution color conversion. However, it has the disadvantage of requiring that more than 8 bits be stored per component color so that interpolated or intermediate image values are interpolated incorrectly during final extrapolation Up to print resolution. It also has the disadvantage of require an extrapolation unit that is capable of producing 1 print-res value interpolated by cycle.

-

                    Interpolate at print resolution, performing at then intensification and color conversion. This has the advantage of a single resampling process, providing maximum precision. However, it has the disadvantage of requiring a unit extrapolation that is capable of producing 1 value bi-cubic interpolated per cycle, as well as perform intensification and color conversion, all on a single cycle average. The intensification function must be large enough to apply the function CAFA-res to the high-res image. Worse, for intensification, at least 3 must be maintained windows on the output image (containing each of them a number of 6,400 input lines) since in a single cycle printing, cyan, magenta and yellow points represent points of 6 different lines.

In none of these cases has it been taken into account the fact that the final print output is bi-level instead of contone. In consecuense, we can print an average primer in relation to resampling, and Get the best of both methods.

The solution is to interpolate to a intermediate resolution Intensification and color conversion occur at intermediate resolution, followed by extrapolation Up to print resolution. The intermediate resolution must be low enough to allow the advantages of Contemplation of the small size intensification function and of the color conversion. But the intermediate resolution must be high enough that there is no loss of extrapolation of Image quality bi-level resolution of Print. The effect must be the same as if there was a simple interpolation to print resolution (instead of two).

Since the print image is printed as 1,600 bi-level dpi points vibrated, it can be represented by a 320 dpi contone image. In consequently, an intermediate resolution of 1,280 pixels contone provides an unperceived loss of quality over 6,400 points bi-level The last increase from 1,280 to 6,400 is therefore an exact extrapolation ratio of 1: 5.

To decide how to perform resampling better, it is better to consider each color plane in relation to the CFA resolution. This has been depicted in Figure 22 for a rotational of 0.

       \ newpage
    

3.2.5.1 Red 45 and Blue 47

Looking at the red 45 and blue 47 planes, the Full version of CFA resolution of the color plane can be created by scaling the number of pixels sampled in each dimension by 2. Intermediate pixels can be generated using a reconstruction filter (such as a Lanczos filter or Exponential). Only one dimension is required in the function, since the function is symmetric. Since the red and the blue they have different deviations in terms of their representation initial within the CFA sample space, the initial positions in the function they will be different.

The mapping of the output coordinates (in the space of 1,280) with respect to the input coordinates, depends on the current rotation of the image, due to the change log of pixels with rotation (either 0 or 90 degrees depending on the print format). For red and blue, it It then maintains the following relationship:

12

where:

x, y = coordinates in the middle res space

x ', y' = coordinates in the space of entry

mps = mean beef pixels in the sample space

k_ {1, \ 2} = {0, -0.5} depending on the rotation.

This means that, given a starting position in the input space, we can generate a new line of medium resolution pixels by adding a \ Deltax and a \ Deltay of 1 / mps and 0, respectively, 1,279 times. The fractional part of x e and in the input space, it can be used directly for the search for function coefficients for construction and resampling of the image.

Note that k_ {1} and k_ {2} are 0 and -0.5, depending on whether the image has been rotated by 0 or 90 degrees. The Table 8 shows the values for k_ {1} and k_ {2} in the planes of the red and blue, assuming that the 90 degree rotation is anti-time

       \ vskip1.000000 \ baselineskip
    

TABLE 8 Rotation effect on k1 and k2 (the rotation is anti-time)

13

The average resolution pixel number per Sample, mps, depends on the print format. Since the image Planarized RGB has the following red and red planar resolutions blue when not rotated: R: 750 x 500, B: 750 x 500, the scalar factors for the different output formats (Figure 13) are shown in Table 9. Note that with the format of Passport image, the entire image is resampled in ¼ of the space output

TABLE 9 Red and Blue Scalar Factors for Formats Image

14

       \ vskip1.000000 \ baselineskip
    

As can be seen in Table 9, the red and blue images are enlarged for all formats of image. Consequently, there will be no overlapping particles introduced by the resampling process.

3.2.5.2 Green 46

Green plane 46 cannot be increased simply in the same way as red or blue, since Each green plane line represents different pixels, either odd or even pixels on alternate lines. Although pixel number terms be representative say the image green is 750 x 1,000, you could also say that the image it is 1,500 x 500. This confusion is due to the nature of chess board of the green pixels, where the distance between pixels is not equal in dimensions x and y, and they do not map well regarding reconstruction or resampling of image. The number of interpolation methods used by other systems for the reconstruction of the green plane, constitutes a testimony regarding all this (from the next replication closest to the linear interpolation until interpolation bi-linear and heuristic reconstruction).

Output coordinate mapping (in space of 1,280) with respect to the input coordinates, is conceptually the same for green as for red and blue. The mapping depends of the current rotation of the image., due to the registration of Pixel changes with rotation (either 0 or 90 degrees depending on the print format). For the green plane it maintains the following relationship:

fifteen

       \ vskip1.000000 \ baselineskip
    

where:

x, y = coordinates in the middle res space

x ', y' = coordinates in the space of entry

mps = mean beef pixels in the sample space

k_ {1, \ 2} = {0, -0.5} depending on the rotation

As with the red 45 and blue 47 planes, the number of average resolution pixels per sample, mps, depends of the print format. Since the planarized RGB image has the following planar resolutions when not rotated: R: 750 x 500; B: 750 x 500, G: 750 x 1,000, the increment factors for the different output formats (see Figure 13) are shown in Table 10. Note that with the Passport image format, the Full image is resampled in ¼ of the output space.

TABLE 10 Scalar Factors of Green Plane for Formats of Image

16

       \ vskip1.000000 \ baselineskip
    

These factors allow coordinate mapping between the CCFA resolution input space and the space of medium resolution However, once we have a coordinate in the CFA resolution input space, we cannot perform the reconstruction and resampling of the image on the samples of the same as in red and blue due to the nature of Flat 46 green chess board.

On the contrary, for the purpose of rebuilding High quality image resampling, we can consider the Green channel is an image rotated in 45 degrees. When we observe the pixels in this way, as Figure 23 shows, appear clearly reconstruction and image resampling methods high quality.

Looking at Figure 23, the distance between pixels sampled in the X and Y directions is now the same. The actual distance between the sampled pixels is \ sqrt {2}, as illustrated in Figure 24.

The solution for the green channel consists then to perform a reconstruction and image resampling in the turned space. Although the same filter is used reconstruction that to resample red and blue, the function It must be different. This is because the relationship between the sampling rate for green and the highest frequency of the signal, it is different from the relationship for the red and blue planes. In addition, the function should be normalized so that \ sqrt {2} of the distance between samples is equal 1 as the function coordinates (non-normalized distances between resampling coordinates should still be used to determine if overlap will occur). Therefore, we need two transformations:

-

                    The first is to map the CFA space not rotated to a rotated CFA space. This can be done by multiplying each coordinate by 1 / \ sqrt {2}, since we are rotating 45 degrees (cos45 = sen45 = 1 / \ sqrt {2}).

-

The second is to increase the coordinates so that they match with the normalized function, which can be done by multiplying each sorted by 1 / \ sqrt {2}.

These two transformations combine to create a multiplication factor of ½. Therefore, according to we advance through the CFA space not rotated x by k, we increase in k / 2 the function x, and we reduce the function y by k / 2. Similarly, as we move in y by k, we increase the function x e by k / 2 we increase the function y / k.

The relationships between two coordinate systems different can be illustrated considering that they occur according to we generate a line of pixels of medium resolution from a input image of CFA space. Given an order of departure in the CFA input space, we start with x = 0, and we advance 1,280 times by 1 / mps, generating a new pixel in each new position. The movement in the CFA space not rotated by 1 / mps, can be decomposed in a movement in x and in a movement in y in the CFA space rotated. The process is shown in Figure 25.

Since cos45 = sen45 = 1 / \ sqrt {2}, the non-rotated movement in CFA space equals equal movement in x and y in 1 / (mps \ sqrt {2}). This amount must be increased. Now to match the normalized function. The increase equivalent to another multiplication by 1 / \ sqrt {2}. Therefore, a movement of 1 / mps in the non-rotated CFA space equals one ½ mps movement in function x and in function y. Table 11 exposes the relationship between the three coordinate systems for the different formats:

TABLE 11 Values Δ of Green Plane Function for Image Formats

17

       \ vskip1.000000 \ baselineskip
    

Table 11 shows that the movement in the function space is always a number less than 1, but in the rotated CFA space, only the Passport image has a \ Delta value greater than 1. As a result, overlap will occur for Passport print format, but not for any of the others. However, since? Is almost 1, and that each of the 4 images is only ¼ size, the overlap will not be appreciable, especially since we assume a low pass filter ideal in green during image capture.

3.2.5.3 Reconstruction Filter for Red, Blue and Green

The exact reconstruction filter that is going to use, will depend on a number of circumstances. Always exists a relationship between the number of samples used between the construction of the original signal, the time invested in the signal reconstruction, and resampled image quality. A satisfactory ratio in this case is 5 pixel samples from the dimension being rebuilt, centered around  of the estimated X position, that is, X-2, X-1, X, X + 1, X + 2. Due to the nature of the reconstruction with 5 sample points, we only need 4 coefficients for the introduction of the convolution function.

We create a coefficient search table of function with n inputs for each color component. Each entry It has 4 coefficients. As we move into the exit space, we map the changes in the output space with respect to the changes in the input space and function space. The most bits significant of the fractional component in the space of the current function, are used for indexing in the table of function coefficients If there are 64 entries in the table of function, the first 6 bits of function are used to search the coefficients, 64 inputs being absolutely sufficient for resampling on the printcam.

3.2.6 Intensification 65

The image captured by the CFA must be intensified before being printed. Ideally, the intensification filter should be applied in the domain of CFA resolution. However, at the resolution of capture of the image, we don't have full color information in every pixel. By On the contrary, we only have red, blue or green in one position of given pixel. The intensification of each color plane shape independent, results in color changes. Intensification it must be applied, on the contrary, to the luminance channel of an image, so that the hue and saturation of a pixel given must remain unchanged.

Intensification then entails translation of an RGB image to a color space in which the luminance is separated from the rest of the color information 80 (such as HLS or Lab). The luminance channel 81 may then be stepped up 82 (with the addition in a proportion of the version filtered high-luminance). Finally, the entire image must be converted back to RGB 83 (or CMY since we are going to print it in CMY). The process is shown in Figure 26.

However, we can avoid many of the stages of color conversion if we consider the effect of adding a filtered high-pass L back to the image; The effect is a change in the luminance of the image. A change in the luminance of a given pixel can be very approximate with an equal change in linear R, G and B. Therefore, we simply generate L, filter passes high L, and apply a proportion of the result in the same way to R,
G and B.

       \ newpage
    

3.2.6.1 Convert RGB to L 80

Consider the CIE 1976 L * a * b color space, where L is significantly uniform. To do the conversion of RGB to L (the luminance channel), we average the minimum and maximum of R, G and B, as follows:

L = \ frac {M \ text {Í} N (R, G, B) + MAX (R, G, B)} {2}

3.2.6.2. High Pass Filter L 84

A high pass filter 84 can be applied Then to the luminance information. Since we are filtering in a medium resolution medium-sized space instead of a CFA resolution space, the function size of intensification can be increased or the result goes high It can be increased appropriately. The exact amount of intensification will depend on the CFA, but a function 85 of 3 x 3 convolution will be enough to produce good results.

If we were to increase the size of the function, Table 12 shows the effective increase required for a 3 x 3 convolution in the CFA space as applied to the 1,280 resolution space, using the green channel as the basis for the increase of the function. From this table it is clear that a function sized at 7 x 7, applied to the space of Medium resolution will be suitable for all intensification.

TABLE 12 Increase Factors for Filter Convolution

18

If a 3 x 3 filter 85 is applied on the med-res image, the result would be increased 86 according to the scale factor used in the operation General image scaling. Given the amounts in Table 12 (particularly, the Standard print format), we can use a 3 x 3 filter 85, and then scale the Outcome. The production procedure of a filtered L pixel Simple is shown in Figure 27.

The actual function used can be a any of a set of standard filter functions passes high.  A basic but satisfactory high pass filter is shown in this PCP implementation in Figure 50.

3.2.6.3 Add L Filtered to RGB

The next thing to do, is in adding some proportion of the filtered luminance values It passes high resulting to the luminance channel. The image can be then converted back to RGB (or, if applicable, CMY). Without However, a luminance change can be reasonably approximated. by an equal change in R, G and B (since the space of color is linear). Therefore, we can avoid all of color conversions by adding an equal amount of filtered luminance value passes high at R, G and B. The exact proportion of the filtered image passes high can be defined by means of a scalar factor.

If L is the filtered luminance pixel passes high, and k is the constant scalar factor, we can define the intensification transformation of R, G and B as follows:

19

Of course, the scalar factor applied to L can be combined with the scalar factor of the filter process pass high (see Section 3.2.6.2) to obtain a single factor climb.

Once the intensification has been applied to the RGB pixel, the image can be converted to CMY 83 in order to Let it be printed.

3.2.7 Conversion to CMY 83

In theoretical terms, the conversion of RGB to CMY It simply consists of:

C

= 1 - R

M

= 1 - G

Y

= 1 - B

However, this conversion assumes that the CMY space has a linear response, which is definitely not it is true in pigmented inks, and it is only partially true in dye-based inks. The individual profile of color of a particular device (input and output), may vary considerably. Therefore, a precise conversion as well as to allow future sensors, inks and printers, it It requires a more accurate model for the Printcam.

The required transformations are shown in Figure 28. Lab is chosen because it is perceptually uniform (unlike with XYZ). In relation to the mapping from the range of Image sensor frequency up to the frequency range of printer, the printer frequency range is typically contained entirely in the sensor frequency range.

Instead of performing these transformations comprehensively, excellent results can be obtained through tri-linear conversion based on 3 sets of 3D search tables. Search tables contain the resulting transformations for the specific input as has indexed by RGB. Three tables are required: a table 90 that maps RGB to C, a table 91 that maps RGB to M, and a table 92 that map RGB to Y. Interpolation can be used tri-linear to provide the final result for those entries that are not included in the tables. The process It is shown in Figure 29.

Tri-linear interpolation it requires reading 8 values from the search table, and carrying out 7 linear interpolations (4 in the first dimension, 2 in the second, and 1 in the third). High precision can be used for intermediate values, although the output value is only 8 bits

The size of the required search table depends on the linearity of the transformations. The recommended size for each table in this application is 17 x 17 x 17 1 , (with every 8 bits of input. A 17 x 17 x 17 table has 4,913 bytes (less than 5 KB).

For indexing in the tables of 17-by-dimension, the components of 8-bit input color are treated as fixed point numbers (4: 4). The four bits of the whole part give the index, and the 4 bits bits of the fraction are used for interpolation.

3.2.8 Ascending Interpolation 67

The medium resolution CMY image (1,280 of width) must now be interpolated ascending to resolution final print (6,400 wide).

The ratio is exactly 1: 5 in both dimensions.

Although it is certainly possible to interpolate bi-linearly the 25 values (1: 5 in both dimensions X and Y), the resulting values will not be printed contone The results will be vibrated and printed. bi-level Since the results of 1,600 dpi contone will be transformed into bi-level points vibrated, interpolation accuracy bi-linear from 320 dpi to 1,600 dpi will not be visible (The average resolution was chosen for this reason). The replication of Pixel will therefore produce good results.

Pixel replication simply entails take a simple pixel, and use it as the value for a large area. In this case, we replicate a simple pixel up to 25 pixels (a 5 x 5 block). If each pixel were contone, the result could appear fractional, but because the pixels are vibrated, the effect is that the resulting 25 bi-level points They adopt the contone value. The process is shown in Figure 30

       \ newpage
    

3.2.9 Halftone 68

Printhead 2 is only capable of Print points bi-level. We must convert, therefore, the image from CMY contone to CMY vibrated. More specifically, we produce an orderly point vibration dispersed using a stochastic vibrating cell, converting a CMY image contone in a bi-level CMY image vibrated

The 8-bit contone value of 1,600 dpi is compare with the current position in vibrating cell 93. If the 8-bit contone value is greater than the vibrating cell value, an output bit that is a 1 is generated. Otherwise, a output bit that is 0. This output bit will be sent eventually to the printhead and control a nozzle simple to produce a simple point C, M or Y. The bit represents if a particular nozzle for a color and a dice position.

The same position in vibrating cell 93 It can be used for C, M and Y. This is because the head of 2 real print produces the points C, M and Y for different lines in the same printing cycle. The staggering of the different colored dots provides an accumulation in the cell vibratory

The halftone process can be appreciated in the Figure 31

The size of the vibrating cell 93 depends on The resolution of the exit points. Since we are producing 1,600 dpi dots, the cell size should be more large 32 x 32. Additionally, to allow the order of point processing matches the head segments of impression, the size of the vibrating cell should be divided ideally for 800 evenly (since there are 800 dots in each segment of the printhead).

A 50 x 50 vibrating cell size is large enough to produce high results quality, and is divided evenly by 800 (16 times). Each entry of the vibrating cell is 8 bits, which gives a total of 2,500 bytes (approximately 1.5 KB).

3.2.10 Reformatting for Printer 69

The final process before being sent to the printer, it is intended for the dots to be formatted in the correct order before being sent to the head of Print. Points must be sent to the printhead in the correct order: 24 points each time as defined in the Section 2.2.1.

If the points can be produced in the order correct for printing (i.e. the vibratory and ascending interpolation generate your data in the correct order), then those point values (each value is 1 bit) can be collected in a simple way, and sent in groups of 24. The process It is shown in Figure 32.

24-bit groups can be sent to continued to printhead 2 by means of the 15 Memjet interface.

4. Core and CPU memory 4.1 CPU core 10

PCP 3 incorporates a single CPU core 10 microcontroller to synchronize image capture chains and print image processing, and to carry out the general operating system tasks of the Printcam, including the user interface. A wide variety of CPU cores are suitable: they can consist of any processor core with sufficient processing power to perform the functions of calculation and control required quickly enough as to meet consumer expectations.

Since all image processing is performed using dedicated hardware, the CPU does not have to process pixels As a result, the CPU can be extremely simple. Without However, it must be fast enough to run the engine step by step during printing (the stepper motor requires a 5 KHz process). As an example of a suitable core, mentioned the Philips 8051 microcontroller that runs at approximately 1 MHz

There is no need to maintain continuity established instruction between different Printcam models. The different PCP chip designs can be manufactured by different manufacturers, without the need to license or perform the CPU core This independence of devices avoids monopoly of the chip provider as has happened in the PC market with Intel

Associated with the CPU Core are a ROM of Program 13 and a small Programmable RAM 14 of Program.

CPU 10 communicates with the other units inside of PCP 3 through memory mapped I / O. The direction particular sets the range map with respect to the units individuals, and within each range, regarding records individuals within that particular unit. This includes the serial and parallel interfaces.

4.2 Program ROM 13

A small Program 13 Flash ROM has been Built-in PCP 3. The size of the ROM depends on the CPU chosen, but will not be larger than 16-32 KB.

4.3 Program RAM 14

Similarly, a small area 14 of RAM programmable is incorporated in PCP 3. Since the Program code does not have to manipulate images, no need of a large work area. The size of the RAM depends on the CPU chosen (for example, stacking mechanisms, convention call sub-routine, sizes of registration, etc.), but should not be greater than about 4 KB.

4.4 CPU Memory Decoder 16

The CPU Memory Decoder 16 is a Simple decoder to satisfy the data accesses of the CPU The Decoder moves the data addresses to the Internal PCP registry accesses by the internal low bus speed, and therefore enables memory mapped I / O of PCP records.

5. Communication interfaces 5.1 USB Serial Port Interface 17

This is a standard USB serial port, connected to bus 18 of low internal chip speed. USB serial port It is controlled by CPU 10. The serial port allows the image transfer to, and from, the Printcam, and allows the DPOF printing (Digital Print Order Format, Digital Order Format Print) of photos transferred under external control.

5.2 QA Chip 19 Series Interface

This consists of two standard serial ports of low speed, connected to bus 18 low speed chip internal. The protocol set to CPU between the two is used to authenticate the print roll [1, 2], and for the following functions:

-

                    Acquire the characteristics of the ink

-

                    Acquire the recommended drop volume

-

                    Drag the amount of printed paper and require new print roll when there is insufficient paper to print the print format required.

The reason for having two ports consists of connect both to Chip 4 QA of connected camera and Chip 5 QA of Printing roll with the use of separate lines. Two chips QA are implemented as Authentication Chips [2]. Yes only one line is used, a replica manufacturer of print rolls could usurp the authentication mechanism [one].

5.2.1 Chip 5 QA of the Printing Roll

Each print roll consumable contains Your own 5 QA chip. The QA chip contains information required for maintain the best possible print quality, and is implemented with the use of an Authentication Chip [2]. 256 bits of data is assigned as follows:

TABLE 13 The 256 bits of the Printing Roll (16)

twenty

twenty-one

       \ vskip1.000000 \ baselineskip
    

Before each print, the amount of paper remaining is checked by means of the CPU to ensure that there is enough for the currently specified print format. Once each print has started, the amount of paper remaining must be decremented on the QA chip of the print roll by means of
CPU

5.3 Parallel Interface 6

Parallel interface 6 connects PCP 3 with individual static electrical signals. The CPU is trained to control each of these connections as mapped I / O of memory via the low speed bus. (See Section 4.4 for more details on memory mapped I / O).

Table 14 shows the connections with the parallel interface

       \ vskip1.000000 \ baselineskip
    

TABLE 14 Connections with the Parallel Interface

22

       \ vskip1.000000 \ baselineskip
    

5.4 7 JTAG interface

A 7 JTAG (Joint Test Action Group) Interface Standard has been included in PCP 3 for verification purposes. Due to the complexity of the chip, a diversity of test techniques, including BIST (Built In Self Test) and functional block insulation. An upper portion of the 10% of the chip area for the global circuitry for checking chip.

6. Image RAM 11

Image RAM 11 is used to store the Image 42 captured. Image RAM is a Flash multi-level (2 bits per cell) so that the image is preserved after feeding has been interrupted

The total amount of memory required for the Planarized linear RGB image is 1,500,000 bytes (approximately  1.5 MB), organized as follows:

-

          R: 750 x 500 = 375,000 bytes

-

                    B: 750 x 500 = 375,000 bytes

-

                    G: 750 x 1,000 = 750,000 bytes.

The image is written by the Capture Unit of Image, and read both by Unit 8 of Image Histogram and Unit 99 Print Generator. CPU 10 does not have access Direct random to image memory. You have to access the Image pixels through the Image Access Unit.

7. Image capture unit 12

The Image Capture Unit contains all the functionality required by the Image Capture Chain, according to It is described in Section 3.1. The Image Capture Unit accepts Pixel data through Image Sensor Interface 98, linearize RGB data through a search table 96, and finally write the linearized RGB image out of RAM in planar format The process is shown in Figure 33.

7.1 Image Sensor Interface 98

Image Sensor Interface 98 (ISI) is a status machine that sends control information to the Sensor CMOS image, including frame synchronization pulses and pulses of pixel clock, in order to read the image. Most the ISI is likely to be used as the manufacturer's source cell of the image sensor. The ISI is controlled in itself by the State Machine 97 of the Capture Unit.

7.1.1 Image Sensor Format

Although a diversity is available of image sensors, we will only consider the device color filter (CFA) Bayer. The CFA Bayer has a number of attributes to be defined now.

The image captured by the CMOS sensor (through of a pickup lens), is supposed to have been sufficiently filtered to remove any overlapping particles. He sensor itself has an aspect ratio of 3: 2, with a resolution of 1,500 x 1,000 samples. The pixel layout plus likely is the Bayer color filter set (CFA), with each 2 x 2 pixel block arranged in 2G mosaic as shown in Figure 15

Each sample contains R, G or B (corresponding to red, green and blue, respectively), is 10 bits Note that each pixel of the mosaic contains information about only one of the R, G or B. Estimates about the missing color information, must be made before for the image to be printed.

The CFA is considered to make some amount fixed pattern noise suppression (FPN). It may require some additional AFN suppression.

7.2. Search Table 96

Search table 96 is a ROM that maps the RGB of the sensor in a linear RGB. This equates process 40 of Linearize RGB described in Section 3.1.2. As such, the ROM is from 3 KBytes (3 x 1,024 x 8 bits). 10 address bits come from the ISI, while 2 bits of TableSelect are generated by the Machine Status 97 of the Image Capture Unit.

7.3 State Machine 97

The State Machine 97 of the Capture Unit Image generates control signals for Sensor Interface 1 Image, and generates directions to linearize the RGB 40 and to planarize image data 41.

Control signals sent to ISI 198 inform the ISI to start pixel capture, finish the pixel capture, etc.

The 2-bit address sent to the Table of Search 96 matches the current line being read from the ISI. For line pairs (0, 2, 4, etc.), the address of 2 bits is Roo, Green, Red, Green, etc. For odd lines (1, 3, 5, etc), the 2-bit address is Green, Blue, Green, Blue. This is true regardless of camera orientation.

The 21-bit address sent to RAM 11 of Image, is the address written for the image. Three records they retain the current address for each of the red planes, green and blue. The increment of the addresses as pixels is Write on each plane.

7.3.1 Records

The Image Capture unit contains a certain number of records

       \ vskip1.000000 \ baselineskip
    

TABLE 15 Records in the Capture Unit of Image

2. 3

Additionally, the Sensor Interface 98 Image contains a number of records. The exact records depend on the Image Sensor 1 chosen.

8. Image access unit 9

Image Access Unit 9 produces the means for the CPU 10 to access the image in ImageRAM 11. The CPU can read pixels from the image of ImageRAM 11 and write pixels again.

The pixels could be read for the purpose of image storage (for example, via USB) 17, or to simple image processing. The pixels could be written in ImageRAM 11 after image processing, as image previously saved (uploaded via USB), or as images for the purpose of test pattern The test patterns could be images Synthetic, specific test images (loaded through the USB), or could be nozzle trigger trip values of 24 bits to be loaded directly into the printhead by middle of the test mode of the Generator Unit 99 Print.

Image Access Unit 9 is a mechanism Direct access to ImageRAM 11, and it operates absolutely simple in terms of 3 records as shown in the Table 16.

TABLE 16 IAU records

24

The structure of the Image Access Unit It is very simple, as shown in Figure 35.

State Machine 101 simply performs read / write from / in ImageRAM 11 whenever the CPU writes in the Mode registration.

9. Image Histogram Unit 8

Unit 8 of Image Histogram (IHU), has been designed to generate histograms of images as required by the Printing Image Processing Chain described in Section 3.2.2. The IHU only generates histograms for images of planar format with samples of 8 bits each.

The Image Histogram Unit 8 is typically used three times per impression. Three different histograms accumulate, one for each color plane. Each time a histogram is accumulated, the results are analyzed in order to determine the low and high thresholds, the scaling factors, etc., for use in the rest of the printing process.
Pressure. For more information on how the histogram should be used, see Section 3.2.2.2 and Section 3.2.4.

9.1 RAM 102 Histogram

The histogram itself is stored in a RAM 102 256 entries, each entry being 20 bits. To the RAM of Histogram is only accessed from within the IHU. Tickets Individuals are read and written as 20-bit quantities.

9.2 State Machine and Records 103

State Machine 103 follows the pseudo-code described in Section 3.2.2.1. The it is controlled by the records shown in the Table 17

TABLE 17 Records in the Histogram Unit of Image

25

26

The typical use of records consists of set TotalPixels with the number of pixels to include in the count (for example, 375,000 for red), StartAddress with the address of the red plane, ClearCount with 1, and write a 1 in the Go record. Once the count is finished, the values Histogram individual can be determined by typing 0-255 on PixelValue and reading the Pixelcount correspondent.

       \ global \ parskip0.900000 \ baselineskip
    

10. Printhead interface 105

The Print Head Interface (PHI) 105, is the means by which PCP 3 loads printhead 2 Memjet with the dots to be printed, and control the Real dot printing process. The PHI is a wrapper logical cyclic for a number of units, namely:

?

        a Memjet Interface (MJI) 15, which transfers data to the Memjet printhead, and controls the nozzle firing sequences during a Print,

?

        a generator unit Printing (PGU) 99 is an implementation of most of the Print Chain described in Section 3.2, as well as It provides a means of producing test patterns. PGU take a planarized linear RGB obtained from an image captured from CFA format from ImageRAM 11, and produces a 1,600 dpi vibrated CMY image in real time as required via Memjet Interface 15. Additionally, the PGU has a Test Pattern mode, which allows CPU 10 to specify from precise way which nozzles have to be fired during a Print.

The units inside the PHI are controlled by a number of records that are programmed by CPU

The internal structure of the Head Interface Print has been shown in Figure 37.

10.1 Memjet interface 15

The Memjet Interface (MJI) 15 connects the PCP with The external Memjet printhead, providing both data as appropriate signals to control the loading sequences and nozzle shot during printing.

The Memjet 15 Interface is simply a Machine of State 106 (see Figure 38) which follows the order of loading and printhead shot described in Section 2.2, and that includes the functionality of the Preheat cycle and Cleaning as described in Section 2.4.1 and 2.4.2.

The MJI 15 loads data in the printhead from an option between 2 data sources:

?

        All 1s. This means that all nozzles will fire during a Print cycle posterior, and constitutes the standard mechanism for loading the head Printing for a Preheating and Cleaning cycle.

?

        From the entrance of 24 bits kept in the Transfer register of the PGU 99. This is the standard means of printing an image, if it is a captured photo or a test pattern. The 24-bit value from the PGU is sent directly to the printhead and a pulse of Advance control of 1 bit to the PGU. At the end of each line, a 1-bit "AdvanceLine" pulse is also sent to the PGU

MJI 15 must be initiated after the PGU 99 has already prepared the first transfer value of 24 bits This is done so that the 24-bit data input will be valid for the first transfer to the head of Print.

The MJI 15 is therefore connected directly to Print Generator Unit 99 and printhead 2 external The basic structure is shown in Figure 38.

10.1.1 Connections to the Print Head

The MJI 15 has the following connections with the printhead 2, with the detection of inputs and outputs with respect to MJI 15. The names are matched with the printhead pin connections (see Section 2).

TABLE 18 Print Head Connections

27

10.1.2 Trigger Pulse Duration

The duration of the firing pulses along the Aunable and BEnable lines depends on the viscosity of the ink (which depends on the temperature and characteristics of the ink), and on the amount of power available for the print head. The typical pulse duration range is 1.3 to 1.8. The MJI therefore contains a programmable pulse duration table, indexed by information from the printhead. The pulse duration table
It allows the use of a lower cost power supply, and helps maintain a more accurate ink ejection.

The Pulse Duration table has 256 entries, and is indexed by the current Vsense and Tsense established. The top 4 bits of the address come from See, and the lower 4 bits of the address come from Tsense Each input is 8 bits, and represents a point value fixed in the range of 0-4 \ mus. The process of Generation of the Aunable and BEnable lines is shown in Figure 39.

The 256 byte table is written by CPU 10 before printing the photo. Each entry of 8-bit pulse duration in the table, combines:

?

        Establishment of glitter

?

        Viscosity curve the ink (from the QA Chip) 5

?

        Rsense

?

        Wsense

?

        Tsense

?

        See

10.1.3 Point Counts

MJI 15 keeps a count of the number of dots of each color shot from the printhead 2. The dot count for each color is a 32-bit value, deleted individually under processor control. Each point count you can maintain a maximum coverage point count of 69 15 cm (6 inch) prints, although in typical use, The point count can be read and deleted after each Print.

While in the initial Printcam product, the consumable contains both paper and ink, it is conceivable that a different Printcam model has a replaceable consumable of ink only. The initial Printcam product can make a countdown of the amount of millimeters left of paper (stored on chip 5 QA, see Section 5.2), to know if there is enough paper available to print the format wanted. There is enough ink for full coverage of the paper supplied. In the alternative Printcam product, the point counts can be used by the CPU to update the 5 QA chip in order to predict when the ink is going to run out of the ink cartridge. The processor knows the ink volume of the cartridge for each of C, M and Y from chip 5. The count the number of drops eliminates the need for ink sensors, and Prevents ink channels from drying out. A drop count Updated is written on the 5 QA chip after each print. A new photo will not be printed unless there is enough ink, and allows the user to change the ink without obtaining a Unsatisfactory photo that has to be reprinted.

The arrangement of the point counter for the cyan, is shown in Figure 40. The 2 point counters remaining (MDotCount and UDotCount, for magenta and yellow respectively), have an identical structure.

10.1.4 Records

The CPU 10 communicates with the MJI through a established record. The registers allow the CPU to parameterize an impression, as well as receiving information about the printing progress

The records that follow are contained in the MJI:

       \ global \ parskip1.000000 \ baselineskip
    

TABLE 19 Memjet Interface Registers

28

29

30

The MJI State Register is a register 16-bit with bit interpretations as follows:

       \ vskip1.000000 \ baselineskip
    

       \ vskip1.000000 \ baselineskip
    

TABLE 20 MJI State Register

31

10.1.5 Preheating and Cleaning Cycles

The Cleaning and Preheating cycles are perform simply by establishing appropriate records:

?

        SetAllNozzles = one

?

        Set the record PulseDuration either in a low duration (in the case of preheating) or in an appropriate ejection duration of drop for cleaning mode.

?

        Set NumLines of so that is the number of times the nozzle should be shot.

?

        Set the Go bit and wait next for the Go bit to be cleared when the Print cycles have been completed.

10.2 Print Generator Unit 99

The Print Generator Unit (PGU) 99 is an implementation of most of the Printing Chain described in Section 3.2, as well as the provision of a Test pattern production medium.

From the simplest point of view, the PGU provides the interface between the image RAM 11 and the Interface Memjet 15, as shown in Figure 41. The PGU takes an RGB linear planar obtained from an image captured from CFA format from ImageRAM, and produces a CMY image 1,600 dpi vibrated in real time as required by the Interface Memjet Additionally, the PGU has a Test Pattern mode, which allows the CPU to specify precisely which nozzles are shoot during a print. The MJI 15 provides the PGU 99 an Advance pulse once all 24 bits have been used, and a pulse AdvanceLine at the end of a line.

The PGU 99 has 2 processing chains of image. The first, the Test Pattern mode, reads simply data directly from Image RAM 11, and format them in a buffer memory ready to provide an output for the MJI. The second contains most of the functions of String Printing (see Section 3.2). The Print Chain shown Figure 18 contains the functions:

?

        Accumulate Statistics 60

?

        Rotate Image 61

?

        White Balance 62

?

        Range Expansion 63

?

        Resampling 64

?

        Intensification 65

?

        Convert to CMY 66

?

        Ascending Interpolation 67

?

        Halftone 68

?

        Reformat for Printer 69.

The PGU 99 contains all these functions with the exception of Accumulate Statistics 60. To perform the stage of Accumulate Statistics, CPU 10 calls the Histogram Unit of Image 8 three times (once per color channel), and apply some simple algorithms The rest of the functions are the domain of the PGU 99 for reasons of precision and speed: precision, because it might be that too much memory is needed to maintain the Full image with high accuracy, and speed, because a single CPU 10 cannot keep high speed demands on Real time print head 2 Memjet.

The PGU 99 takes as input a diversity of parameters, including RGB to CMY conversion tables, constants  to perform white balance and range expansion, Scaling factors for resampling, and access parameters of image that allow rotation.

The two process chains can be seen in the Figure 20. The most direct chain goes from Image RAM 11 to Intermediate Memory 5 through the Access 110 process of Test Pattern The other chain consists of 5 processes, running All of them in parallel. The second process 112 performs the Resampling. The third process 65 performs the intensification. He fourth process 66 performs the color conversion. The 113 process final performs ascending interpolation, semitone and Printer reformat. The processes are connected through of buffers, only with a few bytes between some processes, and with a few kilobytes for others.

Let's look at these processes and buffers mainly in a reverse order, since the timing for the printhead activates the entire process. The timings for particular processes and the requirements of Buffer size are then more obvious. Do not However, buffer sizes are shown as summary in Table 21.

TABLE 21 Buffer sizes for the Unit Print Generator

32

Apart from the number of records, some of the processes have search tables or memory components significant. These are summarized in Table 22.

TABLE 22 Memory requirements within the Processes of the PGU

33

10.2.1 Test Pattern Access

The Test Pattern Access process 110 is the means by which test patterns are produced. Low Normal user circumstances, this process will not be used. It is primarily intended for diagnostic purposes.

       \ global \ parskip0.900000 \ baselineskip
    

Test Pattern Access 110 reads the RAM of Image 11, and pass 8-bit values directly to Memory Intermediate 5 118 to present at the output for the Interface Memjet It does not modify the 8-bit values in any way. The Image 11 RAM data could be produced by CPU 10 using Image Access Unit 9.

Reading data from Image RAM 11 is performs in a very simple cyclic reset way. Be They use two records to describe the test data: start address of the first byte, and the number of bytes. When the end of the data is reached, the data is read again from the beginning.

The structure of the Access Unit 110 Test Pattern is shown in Figure 43.

As can be seen in Figure 43, the Test Pattern Access Unit 110 is little more than the Address Generator 119. When it starts, and with each signal AdvanceLine, the generator reads 3 bytes, produces a pulse TransferWriteEnable, read the next 3 bytes, and then Wait for an Advance pulse. In the Advance pulse, it is provided the Pulse of TransferWriteEnable, the next 3 bytes are read, and the wait occurs again. This continues until the pulse AdvanceLine, with which the process starts again from the current address.

In terms of reading 3 bytes, the Generator Address 119 simply reads three 8-bit values from the Image RAM 11, and writes them in Buffer 5 118. The value of the first 8 bits is written to address 0 of 8 bits of Intermediate Memory 5, the following is written at the address 1 of 8 bits of Intermediate Memory 5, and the third one is written in Address 2 of 8 bits of Intermediate Memory 5. The Generator 110 Direction then waits for an Advance pulse before making the Same thing again.

The addresses generated for Image RAM 11, are based on a start address and a count of bytes as shown in Table 23.

TABLE 23 Pattern Access Records Proof

3. 4

The following pseudo code Illustrates address generation. AdvanceLine and Advance pulses They are not shown.

35

It is the responsibility of CPU 10 to ensure that data have meaning for the printhead 2. Byte 0 is the nozzle trigger data for the 8 cyan segments (bit 0 = segment 0, etc.), Byte 1 is the same for magenta, and the Byte 2 for yellow. Alternate 24-bit sets are intended for odd / even pixels separated by 1 horizontal line of points.

       \ global \ parskip1.000000 \ baselineskip
    

10.2.2 Intermediate Memory 5 118

Buffer 5 118 keeps the points generated from the entire Printing Generation process. Buffer 5 consists of a shift register of 24 bits to keep the generated points, one at a time, from the UHRH 113 (Interpolation Unit Ascending-Semitone and Reformatting), 3 records of 8 bits to keep the data generated from the TPAU (Unit of Test Pattern Access), and a 24-bit register used as buffer to transfer data to the MJI (Interface Memjet). The Advance pulse from the MJI loads the register 24-bit transfer with 24-bit, either from the 3 registers of  8 bits or from the 24-bit single offset register bits

Intermediate Memory 5 therefore acts as double intermediate storage mechanism for points generated, and has the structure shown in Figure 44.

10.2.3 Intermediate Memory 4 117

Buffer 4 117 maintains the image CMY intermediate resolution contone calculated (1,280-res). Buffer 4 is generated through the process 66 of Color Conversion, and is accessible for Interpolation-Ascending process 113, Halftone and Reformatting in order to generate exit points for the printer.

The size of the Contone Buffer, depends on the physical distance between the nozzles of the head of Print. As points of a color are being generated for a physical line, the points of a color are being generated different for a different line. The net effect is that they are printed 6 different physical lines at once from the printer: points odd and even from different output lines, and lines Different by color. This concept has been explained, and the distances have been defined in Section 2.1.1.

The practical conclusion is that there is a distance given in high resolution points from cyan points pairs through magenta points to yellow points odd. In order to minimize the generation of RGB and with it CMY, medium resolution contone pixels are stored temporarily in Intermediate Memory 4.

Since the ratio of average lines Resolution regarding high resolution lines is 1: 5, each Medium resolution line is sampled 5 times in each dimension. TO the effects of buffer lines, we are only affected by those of 1 dimension, so that we will only consider 5 dotted lines from a single pixel line. The distance between nozzles of different colors is 4-8 points (depending on the parameters of the Memjet). Therefore we assume 8, which provides a distance of separation of 16 points, or 17 points in the inclusive distance. He worst case scenario is a scenario in which the line of 17 line points includes the last line of points coming from a given pixel line. This implies 5 pixel lines, with lines of points generated as 1, 5, 5, 5, 1, and allows an increase of nozzle spacing up to 10.

To ensure that the process of generating contone that writes in the buffer does not interfere with the point generation process that is read from memory intermediate, we have added an extra line of medium resolution by color, for a total of 6 lines per color.

The contone buffer has therefore 3 colors per 6 lines, containing each of the 1,280 lines 8-bit contone values. The total memory required is 3 x 6 x 1,280 = 23,040 bytes (22.5 KBytes). Memory only requires a single read of 8 bits per cycle, and a single write of 8 bits for every 25 cycles (each contone pixel is read 25 times). The Intermediate Memory structure 4 is shown in Figure 45.

Buffer 4 can be implemented as a dual-access single cycle RAM (read and write) running at the normal speed of the point generation process printhead, or it can be implemented as running RAM 4% faster with a single read or write access for cycle.

Intermediate Memory 4 is set in the blank (all 0) prior to the start of the process Print.

10.2.4 Ascending Interpolation, Halftone and Reformatting for Printer

Although Interpolation tasks 113 Ascending, Halftone and Reformatting are defined as separate tasks through Section 3.2.8, Section 3.2.9 and Section 3.2.10 respectively, these are implemented as a unique process in the PCP hardware implementation 3.

The entrance to Interpolation Unit 113 Ascending, Semitone and Reformatting (UHRU), is the buffer 117 contone (Intermediate Memory 4) containing the CMY image of Resolution 1,280 (intermediate resolution) previously calculated. The output is a set of 24 bit values in the correct order in to be sent to the Memjet 15 Interface for later print head output via memory Intermediate 5 118. The 24 output bits are generated at a rate of 1 bit each time, and are sent to the 24-bit shift register of Intermediate Memory 5 118.

Control of this process occurs from Advance and AdvanceLine signals from the MJI 15. When UHRU 13 starts, and after each AdvanceLine pulse, it produce 24 bits, and are timed in the register of 24-bit offset of Buffer 5 via a ShiftWriteEnable signal. Once the 24th bit has been timed, a TransferWriteEnable pulse is provided and it is They generate the next 24 bits. After this, the UHRU 113 wait for the Advance pulse from the MJI. When the Advance pulse, the TransferWriteEnable pulse is provided to the Intermediate Memory 5 118, and the next 24 bits are calculated with before a new wait. In practice, once it has been provided the first Advance pulse, synchronization occurs and future Advance pulses will occur every 24 cycles later.

The Ascending Interpolation process, Halftone and Reformatting can be seen in Figure 46.

The Halftone task is undertaken through the Simple comparator 120 without sign of 8 bits. The two entrances to the comparator comes from the Accumulated Vibratory Cell 121 and the Intermediate Report 4 117. The order in which these values are presented to comparator 120 No Sign, is determined by Address Generator 122 State Machine, which ensures that the addresses in the 1,280 resolution image are matched with the oriented segment order required for the head of Print. The Address Generator 122 State Machine therefore undertakes the tasks of Ascending Interpolation and Reformatting for the Printer. Instead of simply accessing a full line every time at high resolution, and then reformat the line according to the printer search requirements (such as has been described in Section 3.2.10), reformatting is achieved by appropriate addressing of the buffer 117 of contone (Intermediate Memory 4), and ensuring that the comparator 120 uses the correct search from cell 121 vibrating to match the accumulated directions.

The Halftone task is the same as it has been described in Section 3.2.9. However, since the outputs of points are generated in the correct order for the head of printing, the size of the Vibrating Cell 121 is chosen so which is divided evenly into 800. Therefore, a position given of the vibrating cell for a segment will be the same for the remaining 7 segments. A 50 x 50 vibrating cell It provides a satisfactory result. As described in the Section 3.2.9, the same position of the vibrating cell can be used for different colors due to the fact that the different lines are being generated at the same time for each One of the colors. The address for the vibrating cell is Therefore very simple. We start in a particular row of the Accumulated Vibratory cell (for example, row 0). The first cell input used is Input 0. We use that input 24 times (24 cycles) to generate 3 colors for the 8 segments, and then we advance to Entrance 1 of row 0. After from Entry 49, we return back to Entry 0. This continues for the entire 19,200 cycles in order to generate 19,200 points. The Halftone Unit is interrupted next and waits by the AdvanceLine pulse that causes the address generator advance to the next row of the vibrating cell.

The 121 Accumulated Vibratory Cell is called so because it differs from a regular vibrating cell by having odd and even lines accumulated. This is because we generate odd and even pixels (starting with pixel 0) over different lines, and Address Generator 122 is prevented from having to advance to the next row and go back over sets 24 pixel alternatives. Figure 25 shows a cell 93 single vibratory, and how to map it in a vibrational 121 cell cumulative of the same size. Note that to determine the "odd condition" of a given position, we number the pixels of a given row as 0, 1, 2, etc.

The 8-bit value from Memory Intermediate 4 117, is compared (unsigned) with the value of 8 bits from Accumulated Vibrating Cell 121. If the value of Pixel of Buffer 4 is greater than or equal to the value of vibrating cell, a "1" bit is set at the output for Intermediate Memory offset record 5 118. In another In this case, a bit "0" is available for the registration of Intermediate Memory offset 5.

In relation to the 19,200 pixels contone of halftone, you must read the 19,200 pixels contone. Unit 122 of Generator of Direction performs this task, generating the addresses in Intermediate Memory 4 117, implementing Effect the UpInterpolate task. The address generation for Reading of Intermediate Memory 4, is slightly more complicated that the generation of direction for the vibrating cell, but not excessively.

The Address Generator to read the Report Intermediate 4, begins only once the first line of the Intermediate Report 4 that has been written. The remaining rows of the Intermediate Memory 4 is 0, so they will be effectively in white (no printed dots).

Each of the 6 effective output lines It has a record with an integer and fractional component. Serving Registry integer is used to select which line of Memory Intermediate should be read for ascending interpolation effective color for those odd and even color pixels particular. 3 anti-pixels are used to maintain the current position within segment 0, and a only temporary counter P_ADR (pixel address) for the deviation to the remaining 7 segments.

In short, address generation for the Intermediate Memory 4 requires the following records, as shown in Table 24.

TABLE 24 Records Required for Intermediate Memory 4 of Reading

36

Initial values for the 6 records of buffer line, are the physical distance between nozzles  (remember that the fractional component is effectively a part divide by 5). For example, if the odd exit points and pairs of a color are separated by a distance of 1 point, and the nozzles of one color are separated from the nozzles of the following by 8 points, the initial values could be as shown in the "First Line" column of Table 25. Once that each set of 19,200 points has been generated, each of these counters must be increased by 1 fractional component, which represents the fact that we are sampling each pixel 5 times in the vertical dimension. The resulting values will then be as shown in the "Second Line" column of Table 25. Note that 5: 4 + 1 = 0: 0 since there are only 6 lines buffer

TABLE 25 Example of Initial and Second Line Structure for the 6 Intermediate Memory Line Registers

37

The 6 intermediate memory line records then determine which of the buffer lines should be read for odd or even pixels of a given color. For determines which of the 1,280 pixels of average resolution should be read from the specific line of Intermediate Memory 4, We use 3 Pixel Direction counters, one for each color, and a single temporary counter (P_ADR) that is used for the Indexing in each segment. Each segment is separated from next for 800 points. In medium resolution pixels, this distance is 160. Since 800 is visible by exactly 5, we only need to use the whole portion of the 3 Pixel Address counters. We generate the 8 addresses for the even cyan pixels, then the 8 directions for the of magenta pairs, and finally the 8 addresses for those of yellow pairs. Next we do the same for the pixels Odd cyan, magenta and yellow. This two set process 24 bits, 24 pairs and then 24 odd ones, it is done 400 times. TO Next we reset the Pixel Address counters (X_P_ADR) to 0, and we advance the 6 memory line registers intermediate. Every 5 line advances, the next line of memory intermediate is now free and ready for update (via the process of Convert to CMY). Table 26 lists the stages of a simple way.

TABLE 26 Address Generation to Read the Report Intermediate 4

38

The pseudo-code to generate the Intermediate Memory addresses 4 117, shown at continuation. Note that it has been related as a set sequential stages. Table 26 shows a better view of the parallel nature of operations during the generation of address.

39

40

10.2.5 Intermediate Memory 3 116

Buffer 3 is a direct set of values R, G, B of 8 bits. These RGB values are pixels. intensified medium resolution (resolution of 1,280), generated through the Intensification process 65, and read by the Process 66 Convert to CMY.

It is not necessary to double Intermediate Memory 116 of temporary storage. This is because process 66 of Read (Convert to CMY) only requires RGB values for the first 39 cycles while the writing process 65 (Intensification) needs 49 cycles before being read to Really update RGB values.

10.2.6 Convert to CMY 66

The conversion of RGB to CMY is done in the medium resolution space (resolution of 1,280) as has been described in Section 3.2.7.

The conversion process 66 must produce the 117 pixels of buffer buffer (Intermediate Memory  4) at a speed fast enough to keep up of Interpolation process 113 Ascending-Halftone-Reformatting. Since each contone value is used for 25 cycles (5 times in each of the dimensions x and y), the conversion process You may need up to 25 cycles. This totals 75 cycles for the set of the 3 color components.

The procedure as described here, requires only 14 cycles per color component, with the RGB input values really released after 39 cycles. Yes the process is implemented with logic that requires access to RGB input values for more than 49 cycles, then the Memory Intermediate 3 116 will require double intermediate storage, since they are updated by process 65 of Intensification after that time.

The conversion is done as interpolation tri-linear Three search tables are used for 17 x 17 x 17 for the conversion process: RGB to Cyan 90, RGB to Magenta 91 and RGB to Yellow 92. However, since we have to perform 25 cycles to perform each interpolation tri-linear, there is no need for any unit of fast tri-linear interpolation. On the contrary, 8 calls to a linear interpolation process 130 is more than suitable.

The address generation for indexing in Search tables, is direct. We use the 4 more bits significant of each 8-bit color component for the address generation, and the least significant 4 bits of each 8-bit color component for interpolation between values retrieved from conversion tables. The address in the search table requires an adder due to the fact that the Search table has dimensions of 17 instead of 16. Fortunately, the multiplication of a 4-bit X number by 17 results in an 8-bit XX number, and therefore does not require no adder or multiplier, and the multiplication of a number of 4 bits per 17 2 (289) is only slightly more complicated, requiring a single addition.

Although interpolation could be performed faster, we use a simple adder to generate addresses and have a single interpolation unit cycle. Therefore, we are trained to calculate the interpolation for the generation of a single color component from the RGB in 14 cycles, as shown in Table 27. The process must be repeated 3 times in order to generate cyan, Magenta and Yellow Faster methods are possible, but it is not necessary.

       \ vskip1.000000 \ baselineskip
    

TABLE 27 Tri-linear interpolation for color conversion

41

As shown in Table 27, you can use a single ADR register and adder for the generation of address in the search tables. 6 sets can be used of 8-bit registers to maintain intermediate results: 2 records keep the values loaded from the tables of search, and 4 records are used for output from the unit of interpolation. Note that the input to the interpolation unit linear always consists of a pair of 8-bit P1 / P2 registers, P3 / P4, and P5 / P6. This is done deliberately to reduce logic of record selection. In cycle 14, the "V" record 131 keeps the 8-bit value finally calculated. The result of 8 bits can be written in the appropriate position in memory Intermediate 4 117 during the next cycle.

A block diagram of process 66 of Conversion to CMY, can be seen in Figure 48.

Assuming the process is executed first to generate cyan, the resulting cyan contone pixel is stored in the cyan contone buffer of resolution 1.280. He process is executed again then on the same entry RGB to generate the magenta pixel. This magenta pixel contone is stores in the magenta resolution buffer contone 1,280. Finally, the yellow contone pixel is generated from the same RGB input, and the resulting yellow pixel is stored in buffer contains yellow resolution 1,280.

The address generation to write in the buffer buffer 117 (Buffer 4), is direct. Only one address is used (and the ColorSelect bits that the accompany) to write in each of the three memories color intermediates Cyan's buffer is written in the cycle 15, Magenta in cycle 30, and Yellow in cycle 45. The pixel direction is increased by 1 every 75 cycles (after that all 3 colors have been written). The line in the one that is being written increases with restart once every 5 pulses of AdvanceLine. The order of the lines that are being writing is simply 0-1-2-3-4-5-0-1-2-3 ... etc. That way, the scriptures (25 x 1,280 x 3) are balanced with the readings (19,200 x 5).

10.2.7 Intermediate Memory 2 115

Intermediate Memory 2 accepts the output from process 112 of Resimple-CreateLuminance, where a pixel is generated RGB and full L for a given pixel coordinate. The exit of the Intermediate Memory 2 115 goes to Intensification process 65, which requires a set of 3 x 3 of 135 luminance values centered on the pixel that is being intensified.

Therefore, during process 66 of intensification, there is a need for 3 x 3 matrix access luminance values, as well as 136 RGB value corresponding for the central luminance pixel. The same time, the following 3 luminance values and the central value Corresponding RGB must be calculated using process 112 of Resimple-CreateLuminance. The logical view of Access to Intermediate Memory 2 115 is shown in Figure 49.

The actual implementation of the Intermediate Report 2 115 is performed simply as a 4 x 6 RAM (24 inputs) of 8 bits, providing addressing over read and write the effective displacement of values. A column counter of 2 bits, can be increased with reboot to provide a cyclic buffer, which effectively implements the equivalent offset of buffer data Complete in 1 column position. The fact that we don't have need of the fourth column of RGB data, is not relevant, and simply use 3 bytes and save from not having to implement no displacement or complicated read / write logic. In a given cycle, RAM can either be written or be read. The reading and writing processes have 75 cycles to complete in order to keep pace with the head of Print.

10.2.8 Intensification

Intensification unit 65 performs the task of intensification described in Section 3.2.6. Market Stall that intensified RGB pixels are stored in Memory Intermediate 3 116, Intensification Unit 65 must follow the rhythm of the process 66 of Converting to CMY, which implies that a Full RGB pixel must be intensified within 745 cycles

The intensification process includes a filter L-pass (a channel generated from the RGB data and stored in Buffer 2), and add the L filtered back to RGB components, as described in the Table 12 within Section 3.2.6.2. The filter high pass used is a filter basic high pass that uses a function of 3 x 3 convolution, as shown in Figure 50.

The high pass filter is calculated Over 10 cycles The first cycle loads the temporary register 140 with 8 times the central pixel value (the central pixel shifted to the left in 3 bits). The next 8 cycles subtract the 8 values of pixel remaining, with a background of 0. That way, the procedure Complete can be done by an adder. Cycle 10 it includes the multiplication of the result by a constant 141. This constant is the representation of 1/9, but it is a record that allows the amount to be altered by software, by Some scalar factor.

The total amount is then added to the R, G and B values (with a ceiling of 255), and are written in Memory Intermediate 3 during cycles 72, 73 and 74. Calculate / write the RGB values intensified during the last 3 cycles of the set of 75 cycles, eliminates the need for a double Intermediate temporary storage in Intermediate Memory 3.

The structure of the Intensification unit It can be seen in Figure 51.

The adder unit 142 connected to the Memory Intermedia 2 115, is a subtractor with a bottom of 0. The TMP 140 the first value of L is loaded with 8 x during cycle 0 (of 75), and then the next 8 values of L are subtracted from it. The result is not signed, since the subtraction has as background 0.

During the 10th cycle (Cycle 9), the total of 11 bits of the TMP 140 are multiplied by a scalar actor (typically 1/9, but under software control, so that the factor can be set), and rewritten in TMP 140. Only 8 bits integers of the result are written in the TMP (the fraction is discard), so the unit limit of multiplying is 255. Yes a 1/9 scalar factor is used, the maximum written value will be 226 (255 x 8/9). The scalar factor is 8 bits of fraction, with the bit representing 1/8. The variable scalar factor can take into account the fact that different print formats are the result of scale the CFA image with a different amount (and thus, the 3 x 3 convolution will produce correspondingly results scaled).

The intensified values for red, the green and blue, are calculated during Cycle 72, Cycle 73 and the Cycle 74, and are written in the R, G and B records of the Report Intermediate 3 116, one write per cycle. The calculation made in these 3 cycles simply consist of the addition of TMP to R, G and B of Intermediate Memory 2 corresponding to the central pixel.

The address generation is direct. to write in Intermediate Memory 3 116 is simply R, G and B in the cycles 72, 73 and 74, respectively. Reading in memory Intermediate 2 115 makes use of the cyclical nature of Memory 2 Intermediate. The address consists of a column component of 2 bits (which represents which of the 4 columns should be read), and a 3-bit value representing L1, L2, L3, R, G or B. The number of column starts with 1 on each line, and increases (with restart cyclic) every 75 cycles. The order of reading in the Memory Intermediate 2 is shown in Table 28. Record C is a 2-bit column component of the address. Any addition on C is module 4 (cyclically restarts within 2 bits).

       \ vskip1.000000 \ baselineskip
    

TABLE 28 Reading Access to Intermediate Memory 2 during the set of 75 cycles

42

After Cycle 74, register C retains the column number for the next calculation set, making so the search during the next Cycle 0 is valid.

Intensification can only begin when there are enough L and RGB pixels written in the Memory Intermediate 2 (so that the high pass filter is valid). The intensification process must therefore stop until the process of writing Intermediate Memory 2 has advanced 3 columns.

10.2.9 Intermediate Memory 1 114

Buffer 1 maintains the pixels of balanced white and expanded range at the spatial resolution of original capture Each pixel is stored with 10 bits of resolution color, compared to the resolution of 8 bits per pixel of RAM image storage color resolution of image.

Intermediate Memory 1 is arranged as 3 separately addressable buffers: one for each color plane of red 145, green 146 and blue 147. A simple summary of the buffers in Figure 52.

During the course of the 75 cycles, they are read 16 entries from each of the 3 buffers 3 times through the resampling process 112, and up to 29 are written new values in the 3 buffers (the exact number it depends on the scalar factor and the current position of sub-pixel during resampling).

Buffers must be wide enough for reading and writing you can perform without interfering with each other. During the process of reading, 4 pixels are read from each of the 6 rows. If he scalar factor is very large (for example, we scale to Panoramic), the same input pixels can be read multiple times (using a different function position for resampling). Eventually, however, the following pixels will be required. If we do not scale so much, new pixels may be required. before the next pixel generation cycle (that is, within of the 75 clock cycles).

Observing the scalar factors of Table 9 and from Table 11, the worst case for scaling is format 31 Passport:

?

        The green plane has a Δ value for Passport of 1.5625, indicating that 4 positions can be contained within 6 pixel positions CFA However, each row of green samples only maintains Each alternate pixel. This means that only 4 are needed. samples per row (the worst case is 4, not 3, due to the worst case of initial position). The Y-movement indicates the requirement of a additional sample column, which does 5. Finally, it is required An additional sample column for writing. This gives a total of 6 samples per row. 7 rows are required for a sample simple. To generate the 3 sets of RGB pixels for each position x, the maximum movement in y will be 4 rows (3,125 = 2 x 1.5625). The X movement adds a sample row above and by down. Therefore, a total of 13 rows is required. For more details, see Section 10.2.10.

?

        The red and blue planes they have a Δ Passport value of 0.78125, which indicates that 4 positions can be contained inside 4 samples. An additional sample is required to write while the 4 Remaining are being read. This gives a total of 5 samples per row, which is also increased up to 6 samples to match with the green plane (for start-up purposes). 6 rows are required to provide movement in and. For more details, see Section 10.2.10.

Each intermediate sub-memory is implements as a RAM with decoding to read or write a Simple sample of 10 bits per cycle. The Buffers have been summarized in the Table 29, and consume less than 200 bytes.

       \ vskip1.000000 \ baselineskip
    

TABLE 29 Sub-Memory Summary Intermediate

43

10.2.10 Resample and Create Luminance Channel

The 112 Resample and Create Channel Process 112 Luminance is responsible for generating RGB pixel value in a medium resolution space by appropriate resampling of the R, G and B planar images of expanded range and white balanced, as described in Section 3.2.5. Further, luminance values for the given RGB pixel, as well as luminance values for the pixel above and below the RGB pixel, must be generated for use in the last process of intensification.

The time allowed to produce the RGB value and 3 L values is 75 cycles. Since L is simply the average of minimum and maximum of R, G and B for a given pixel position (see Section 3.2.6.1), we must effectively produce values RGB for 3 pixel coordinates: the pixel in question and the pixel of above and below. Thus, we have 75 cycles in which calculate the 3 medium resolution RGB samples and their L values corresponding.

Enter L values (and therefore RGB values) to save the recalculation, it requires too much memory, and in any case, we have enough time to generate RGB values. Buffer 4 117 contains pixels of medium resolution, but cannot be used because contains intensified CMY pixels (instead of RGB pixels not intensified).

10.2.10.1 Resampling

The resampling process can be seen as 3 RGB generation sets, each of which must be completed within 25 cycles (for a maximum total time after 75 cycles). The process of generating a value Simple RGB can be seen, in turn, as 3 processes performed in parallel: the calculation of R, the calculation of G, and the calculation of B, all this for a pixel coordinate of given average resolution. The theory to generate each of these values, can be found in Section 3.2.5, but the conclusion is to execute effectively three image reconstruction filters, one in each Image channel In the case of the PCP, we perform the image reconstruction with 5 sampling points, requiring 4 coefficients in the convolution function (since a coefficient is always 0 and thus the point of sample).

Therefore, the calculation of pixel R of medium resolution is achieved by running a rebuild filter of image on the data R. The calculation of the pixel G of resolution media is achieved by running an image resolution filter over G data, and the average resolution pixel B calculation is achieved running an image reconstruction filter on the B data. Although the functions are symmetric in x and y, they are not the same for Each color plane. R and B are probably the same function because to its similar image characteristics, but the G plane, because to the rotation required for image reconstruction, you must have A different function. The high level view of the process can be appreciated in Figure 53. The address generation has not been shown.

The resampling process can begin only when there are enough pixels in Memory Intermediate 1 for the current pixel line that is being generated. This will be the case once 4 columns of data have been written in each of the color planes in Intermediate Memory 1 114. Resampling process 112 must stop until that moment.

To calculate a resolution pixel value Average of a given color plane, we have 25 cycles available. To apply the 4x4 sample area function, we apply the function 1D (indexed by x) on each of the 4 rows of 4 input samples. Next we apply the 1D function (indexed by y) on the 4 resulting pixel values. He Final result is the resampled output pixel. The application of  a single coefficient for each cycle provides a total of 16 cycles to generate the intermediate r values, and 4 cycles to generate the final pixel value, for a total of 20 cycles.

With respect to accuracy, the pixels of input are 10 bits each (8: 2), and the coefficients of the function are 12 bits. We maintain 14 bits of precision during the 4 stages of each application of the function (8: 6), but We only save 10 bits for the result (8: 2). So, you can use the same convolution wit when the convolution is perform in x and y. The final output of R, G or B is 8 bits.

The heart of the resampling process what It constitutes Convolution Unit 50, as shown in Figure 54

The resampling process then includes 20 cycles, as shown in Table 30. Note that Row 1, Pixel 1, etc., refers to the entry from the Intermediate Memory 1 114, and must be taken into consideration by the mechanism of addressing (see below).

TABLE 30 20 Cycle Resampling

       \ vskip1.000000 \ baselineskip
    

44

       \ vskip1.000000 \ baselineskip
    

10.2.10.2 Generation of L 8-

As described in Section 3.2.6.1, we must convert 80 from RGB to L for the subsequent intensification process. We consider the color space L * a * b CIE 1976, where L is perceptually uniform. For the conversion from RGB to L (the luminance channel), we average the minimum and maximum of R, G and B, as
follow:

       \ vskip1.000000 \ baselineskip
    

L = \ frac {M \ text {Í} N (R, G, B) + MAX (R, G, B)} {2}

       \ vskip1.000000 \ baselineskip
    

The generation of R, G and B values of a pixel given, it is done in parallel, taking 20 cycles. Total time For the generation of L as described here, it is 4 cycles. This Makes the total generation time of an RGBL pixel set be 24 cycles, with 1 cycle to share (since the process must be completed within 25 cycles).

The value of L can be written like this safe in Intermediate Memory 2 115 in the 25th cycle. The generation  Address is described below.

A simple 8-bit comparator can produce 3 bits in 3 cycles, which can be used later to select the 2 entries for the adder, as shown in the Table 31. The division by 2 can be incorporated simply in the adder.

TABLE 31 Selection of Min and Max based on 3 comparisons

46

Since the adder simply adds the value minimum to maximum, the order is not important. Therefore of the 2 inputs to the adder, Input1 may be an option between R and G, while Input2 is an option between G and B. Logic is a minimization of appropriate bit patterns in Table 31.

10.2.10.3 Address Generation for Memory Intermediate 2

The output from the resampler is an RGB pixel simple, and 3 pixels of luminance (L) centered vertically on the RGB pixel The 3 values of L can be written in the Memory Intermediate 2, one every 25 cycles. R, G and B values must be written after cycle 45 and before cycle 50, since the second pixel generated is the central pixel whose RGB values must be maintained The Buffer2 address consists of a component of 2-bit column (which represents which of the 4 columns should be written), and a 3-bit value representing L1, L2, L3, R, G, or B. The column number starts at 0 on each line, and increases (with cyclic restart) every 75 cycles (that is, after write L3).

10.2.10.4 Address Generation for Search Function

The method of calculating the address of the function is the same as described at the end of Section 3.2.5.  Each function is 1 dimension, with 64 entries in the table. The 6 most significant (truncated) bits of fractional component in the Current function space, are used to index in the table of function coefficients For the first 16 cycles, the ordinate X It is used to index the function, while in the following The ordinate Y is used for 4 cycles. Since the function is symmetrically, the same function can be used for both X and Y.

For each of the 1,280 values resampled, we need to produce 3 pixels: the pixel in question 161, and pixels above 160 and below 162 of that pixel. Instead of generating a central pixel and then moving up and down from the central pixel, we generate a pixel 160 and generate two pixels 161, 162 below it. The second pixel 161 generated the central pixel is taken. Then we return to the row  original and we generate the next 3 pixels in the following starting position In this way, as Figure 55 shows, We generate 3 pixels for each of the 1,280 positions.

That way, we have a current position in the space of the function. As we move on to the next pixel in X or Y in the original input space, we add values of appropriate data to these function coordinates. Observing the Figure 56, we see two cases of rotated entry space and not turned.

Consider that the movement in X and Y as \ DeltaX and \ DeltaY, their values being format dependent printing, and therefore the value of mps (see Section 3.2.5). For the green channel, ΔX = ΔY = ½ mps. For the Red and blue channels, ΔX = 1 / mps and ΔY = 0. See Table 9 and Table 11 for appropriate values of ΔX and of ΔY.

Now we can apply the values \ DeltaX and ΔY to movement within the function. Therefore, when we advance in X, we add \ DeltaX to X and subtract \ DeltaY of Y. In the case not rotated, simply 0 is subtracted from Y. mode, when advanced in Y, we add \ DeltaY to X and \ DeltaX to Y. We can do this because the movement in X and Y differs in 90 degrees.

The address generation for the search for function assumes a starting position established by software, and two deltas \ DeltaX and \ DeltaY with respect to Y movement in the function space. The logic of function generation is shown in pseudo-code that follow:

47

As shown in the pseudo-code, the 3-pixel generation occurs 1,280 times. Associated with the generation of each pixel, there are 2 additions, which can be made during the course of the 25 cycle task GeneratePixel. Each GeneratePixel task is from 25 cycles, consisting of 4 sets of 4 cycles that index the function through KernelX (coefficients 0, 1, 2, 3), followed by 4 cycles that index the function through KernelY (coefficients 0, 1, 2, 3), followed by 9 standby cycles.

Note that all values are positive and Only fractional. The two embodiments from the update of the X and Y values of the function, are arranged at output for address generation of the Buffer 1 (see Section 10.2.10.5). These banners of realization simply indicate whether the particular ordinates for the function were cyclically restarted or not during the mathematical operation. The cyclic reset can be either above 1 or by below 0, but the result is always positive.

The two realization bits are always sent to the Rotate Unit / White Balance / Range Expansion, to its use in determining relative input lines of the image.

10.2.10.5 Address Generation for Memory Intermediate 1

Resampler 112 reads from Memory Intermediate 1 114, which you sewed in 3 buffers 145, 146, 147 individually addressable: one for each color plane. Each buffer can be either read or written, during each cycle

The 75 cycle reading process is divided in 3 sets of 25 cycles: a set of 25 cycles for the generation of each pixel. Each set of 25 cycles includes 16 Intermediate Memory 1 readings, followed by 9 cycles without no access Intermediate Memory 1 is written during these 9 cycles The 16 readings of Intermediate Memory 1 114, are effectively 4 sets of 4 readings, and match 4 groups of 4 readings in the function for each color plane.

The address generation then includes the 16-way generation to calculate the first pixel (followed 9 standby cycles), generating 16 addresses for calculate the second pixel (followed by 9 standby cycles), and finally generating the 16 addresses for the third pixel (followed by 9 standby cycles).

Each color plane has its own parameters Address of Intermediate Memory 1 starting. As they are generated the 3 sets of addresses for each of the 1,280 positions along the line, and as the sampler advances from a line of 1,280 samples to the next, the two bits of implementation from the Direction Generation Unit Function are used to update these parameters of Intermediate Memory address 1.

10.2.10.6 Green Buffer 146

The address generation for the intermediate sub-memory 146 of the green inside of Intermediate Memory 1 114, is more complicated than in the red buffer 145 of the red and in the sub-buffer 147 of blue, for two main reasons:

?

        the green channel represents a pattern as a chess board in the CFA. The Alternate lines consist of odd or even pixels only. For resample the green channel, we must effectively turn the 45 degree channel.

?

        there is twice as much green pixels than red or blue pixels. Resample means reading more samples in the same amount of time: there are still 16 samples read to generate each pixel in the space of medium resolution, but there is a higher probability of doing advance the buffer every time. The exact probability It depends on the scalar factor used.

However, the same concept of use of a RAM as a cyclic buffer for the green channel. The green sub-buffer it is a RAM of 78 entries with a logical structure of 13 rows, each containing 6 entries. The relationship between RAM address and logical position is shown in Figure 57.

Samples in Intermediate Memory 1 146 they represent a pattern as a chess board in the CFA. In consequently, the samples in a row (for example, the addresses 0, 13, 26, 39, 52, 65) can represent odd or even pixels, depending on the current line within the entire image, and if the Image has been rotated or not by 90 degrees. This is illustrated in the Figure 58

Therefore, when we map an area of 4x4 sample in the buffer, there are two possibilities  for the interpretation of the samples. As a result, there are two types of addressing, depending on whether the current line is represented by odd or even pixels. This means that the even rows with image rotation 0 have the same addressing that odd rows with image rotation 90, since both contain odd pixels. Similarly, the odd rows with image rotation 0 will have the same addressing that even rows with image rotation 90, since both contain even pixels. The decision has been summarized in Table 32.

       \ vskip1.000000 \ baselineskip
    

TABLE 32 Sample Type Determination

49

The 4 x 4 rea sampling window is the way in which we effectively rotate the buffer by 45 degrees. The 45 degree rotation is necessary for effective resampling, such as It is described in Section 3.2.5.

Assuming for the moment that only we need to generate a simple resampling, we consider the addressing the buffer memory by examining two Types of 4x4 sampling windows as shown in Figure 59.

Although the two types of 4x4 sampling seem similar, the difference is derived from the way it is represented 4x4 mapping in the planar image. Figure 60 illustrates the mapping of 4x4 sampling of Type 1 in the sub-memory intermediate green. Only the top 7 and 4 rows rightmost columns have been represented, since the 4x4 sample area is completely contained within this area.

Buffer pixel mapping for sample the rows for the Type 2 sampling process, it is very similar and can be seen in Figure 61.

In both Type 1 and Type 2 addressing The 16 samples have two ways of processing a row. He Addressing processing of Rows 1 and 3 of Type 1 is the same (relatively speaking) as row processing Type 2 and 3 2. Similarly, processing rows 2 and 4 Type 1 is the same (relatively speaking) as the processing of rows 1 and 3 of Type 2. We will name these methods of Type A 170 and Type B 171 row addressing, as shown in Figure 62.

Given a starting position for the 4 x window 4 (WindowStartAdr) and a type of heading (WindowStartType), we can generate the addresses for the 16 samples by means of a table of 8 inputs (to pass through the two sets of 4 samples). When we read the first sample value, we add a deviation of the table to get to the next position. Deviation It will depend on the type (A, B = 0, 1). The deviation from the fourth Sample is the amount needed to reach the first point of Sample for the next line (and you should consider the number of sample columns). After generating each row of 4 samples, we exchange Type A and Type B. The logic to generate the addresses for a simple set of 16 samples, appears in the Pseudo-code that follows. The add module 78 Provides cyclic storage.

fifty

The search table consists of 8 entries: 4 for the generation of address deviation of Type A 170, and 4 for Type B 171. The deviations are all relative to the Current sample position (Adr).

       \ vskip1.000000 \ baselineskip
    

TABLE 33 Deviation Values for Address Generation of 16 samples

51

At the end of the 16 readings, the TypeAB bit will be the same as the original value (loaded from WindowStartType).

Reading a single set of 16 samples is not enough. Three sets of 16 samples should be read (which represent 3 different positions in Y in the input space not rotated). At the end of the first and second sets of 16 samples, the positions of the function are updated by the generator of function address. The realization bits of this update are used to set the window for the Next set of 16 samples. The two realization bits are indexed in the table that contains a deviation and a banner of 1 bit The deviation is added to WindowStartAdr, and the banner is Use to determine whether or not WindowStartType is reversed. The values for the table are shown in Table 34.

TABLE 34 WindowStartAdr update and WindowStartType

52

At the end of the third set of 16 samples, the function positions are updated to compensate in advance in X in the input space not rotated. This time, there is a different direction of movement, so a table is used Offset / TypeAB modification different. We can't add these deviations to the current WindowStartAdr value, because represents a position of two Y-movements out of Where we want to start the movement. Consequently, we charge WindowStartAdr and WindowStartType from another set of variables: TopStartAdr and TopStartAdr, which represent the first entry in the current line of 1,280. The two banners of realization from the address generator of the Function, are used for the search in Table 35, to determine the deviation that It has to be added to TopStartAdr and whether or not TopStartAdr is invested. As before, the addition is module 78 (the RAM size of the green). The results are copied to WindowStartAdr and WindowStartType for use in the generation of the following 3 sets of 16 samples.

TABLE 35 TopStartArt update and TopStartType

53

After the processing of the 1,280 sets of 3 sets of 16 samples, the next line of 1,280 begins. However, the address of the first sample for position 0 within the next line must be determined. Since the Samples are always loaded in the correct places in the Memory Intermediate 1, we can always start exactly from it position in Buffer 1 (i.e., TopStartAdr can be loaded from a constant Position0Adr). However, we must worry about what type we are dealing with, since the type It depends on how much we advance. Therefore, we have a Initial Position0Type that must be updated depending on the banners of realization from the address generator of function. Since we are moving in the entrance space AND not turned on, the logic used is the same as for the update of the WindowStartType, except that it is performed on Position0Type. The new value for Position0Type is copied to TopStartType, and in WindowStartAdr to start sampling the First position of the new line.

The sampling process for a line of 1,280 position given, cannot start until there are entries sufficient in Intermediate Memory 1, located there by the Unit Turn / White Balance / Range Expansion. This will happen 128 cycles after each new line (see Section 10.2.11).

10.2.10.7 Red and Blue Buffers

Buffer sub-memory 145 of red and intermediate sub-memory 147 of blue Intermediate Memory 1, are simply 2 RAMs that are accessed as cyclic buffers. Each buffer is of 30 bytes, but it has a logical organization of 6 rows, each of which contains 6 entries. The relationship between the address of RAM and the logical position, is shown in Figure 63.

For red and blue, the first 16 samples to be read are always the 4 x 4 top entries. Do not the two columns of remaining samples can be accessed by part of the reading algorithm in this phase.

The address generation for these first 16 samples is simply a starting position (in this case 0) followed by 16 stages of addition module 36, as shown in the following pseudo-code:

54

However, this mechanism for generating address is different from the green channel. Instead of designing two steering mechanism, it is possible to apply the scheme of addressing from green to red and blue channels, and Simply use different values in the tables. This reduces The complexity of the design. The only difference then, lies in the add module 36, instead of the add module 78. This can  Be provided by a simple multiplexer.

Viewing the various address tables for the green, and considering them as applied to red and blue, it is Obviously there is no need for a Type, since both red and blue channels do not need to be turned 45 degrees. That way, we can safely ignore the Type value. The red / blue equivalent of Table 33, shown in Table 36, It has two sets of 4 identical inputs.

TABLE 36 Deviation Values for Address Generation 16 Samples (Red / Blue)

55

As with the address generation of the green, we move twice in Y before moving on to the Next entry of 1,280. For red and blue there is no scaling between the movement in the function space and the movement give the entrance space. There is also none rotation. As we move in Y, the \ DeltaY of 0 is added to KernelX (see generating function address in Section 10.2.10.4). As a result, the realization of KernelX Looking at Table 34, the only possible occurrences are the values 00 or 01 KernelX / KernelY. In the case of 00, the solution green is not to change to any of WindowStartAdr nor WindowStartType, so this is also correct for red and the blue In the case of 01, we want to add 1 to WindowStartAdr, and there is no need to be careful about WindowStartType. The green values can therefore be used so Safe for red and blue. The worst case is the advance in 1 in the address both times, resulting in a worse case of overlap as shown in Figure 65.

At the end of the third set of 16 samples, TopStartAdr and TopStartType must be updated. Since we we are moving in X (and adding \ DeltaY = 0 to KernelY), the realization from KernelY will always be 0. The equivalent of red / Table 35 blue is shown here in Table 37. Note that there is no Type column, since Type is not important for Red or Blue.

TABLE 37 Update of TopStartAdr and TopStartType (Red / Blue)

56

The advancement process from a 1,280 line sets of 3 pixels until the next one, is the same as for the green. The Position0Adr will be the same for the first set of 16 samples for a given line (Position0Adr = 0 for red and blue), and Type is irrelevant. The red and blue generation must start at the same time as the green generation, so that cannot start up to 128 cycles after the start of a new one line (see Section 10.2.11).

10.2.11 Turn, White Balance and Range Expansion 111

The real task of loading the Intermediate Memory 1 114 from image RAM 111, includes the rotation stages, white balance, and expansion of rank 111, as described in Section 3.2.3 and Section 3.2.4. The pixels must be produced for Buffer 1 fast enough for use by the Resampling process 112. This means that during a simple group of 75 cycles, this unit must be capable of reading, processing, and storing 6 red pixels, 6 Blue pixels and 13 green pixels.

The optional rotation stage is undertaken with the pixel reading in the appropriate order. Once a pixel given has been read from the appropriate plane in the storage of the image must be balanced on the white and its adjusted value of according to the range expansion calculation defined in the Section 3.2.4. The process simply includes a single subtraction (background 0), and a multiplication (ceiling 255), both versus constants color specific The structure of this unit is shown in the Figure 66

The low thresholds 72 of red, green and blue, together with the 173 scalar factors of red, green and blue, it determined by CPU 10 after generating the histogram to each color plane through Unit 8 of Histogram of Image (see Section 9).

Depending on whether the current pixel is being processed on the line is red, green or blue, they are multiplexed the appropriate low threshold and scalar factor in the unit of subtraction and in the multiplication unit, with the output written in the appropriate color plane in the Intermediate Memory one.

Subtraction unit 172 subtracts the value Low threshold of 8 bits from the pixel value of Image RAM 8 bits, and has a background of 0. The result of 8 bits is made pass to the specialized multiplication unit 8 x 8, the which multiplies the value of 8 bits by the scalar factor of 8 bits (8 fraction bits, integer = 1). Only the Top 10 bits of the result, and represent 8 bits of the part integer and 2 bits of the fraction. The multiplier 174 has a roof 255 result, so if any bit greater than 7 bits would have been established as a result of multiplication, the result of the entire 8-bit integer is set to 1s, and the fractional part is set to 0.

Apart from the subtraction unit 172 and the unit 174 of multiplication, most of the work of this unit is done using Address Generator 175, which effectively constitutes the state machine for the unit. The address generation is governed by two factors: in a cycle given, you can only access the Image RAM 11, and in a given cycle, only one access to the Intermediate Memory 1 114. Of the 75 cycles available, they are used 3 sets of 16 cycles for reading the Intermediate Memory 1. The actual use is 3 sets of 25 cycles, with 16 readings followed by 9 standby cycles. That gives a total of 27 cycles available for 25 writings 86 red, 6 blue, 6 green). This means that the two restrictions are satisfied if the Script timing in Intermediate Memory 1 matches with the waiting cycles of the resampler 112.

10.2.11.1 Address Generation for Memory Intermediate 1

Once the process is running resampling, we are only affected by writing in the Intermediate Memory 1 during the period in which Resampler 112 He is not reading in it. Since the resampler has 3 sets of 16 readings every period of 75 cycles, there are 27 cycles available for writing. When the resampler does not is working, we want to load Intermediate Memory 1 so fast as possible, which means a write in memory Intermediate 1 114 each cycle. The Generation of Address for the Buffer 1 runs out of a state machine that Take these two cases into consideration. Whenever a value from to Image RAM 111, the adjusted value is written in the appropriate color in Intermediate Memory 1 one more cycle late.

The Address Generation for Memory Intermediate 1 therefore includes a single address counter for each of the sub-memories of the Red, blue and green. The initial address for RedAdr, BlueAdr and GreenAdr, is 0 at the beginning of each line in each case, and then of each write in Intermediate Memory 1, the address is Increase by 1, with cyclic restart by 36 or 78, depending on whether red, green or red are being written to the buffer blue. Not all colors are written every 75 cycle period. A column of green will typically require refilling to double speed than red or blue, for example.

The logic is shown in the Pseudo-code that follows:

       \ vskip1.000000 \ baselineskip
    

57

58

10.2.11.2 Address Generation for RAM of Image

Each map can be read in one of two orientations: rotated in 0 or 90 degrees (turn anti-schedule) This effectively translates as a row mode or column mode access to the planar image. Additionally, we allow marginal pixel replication or constant color for readings outside the limits of the image, as well as the cyclic restart of the image for formats of Printing such as Passport 31.

At the beginning of each print line we must read ImageRAM 11 to load Buffer1 114 as fast as it is possible. This amounts to a simple access to a sample of each cycle. Resampling can only occur once they have been loaded 5 columns, which means 5 columns of 6, 6, and 13 samples, for a total of 125 cycles. In addition to an extra cycle for the final value to be written in Buffer1 114 after have been loaded from ImageRAM 11. To make counting easier, We round up to 128 cycles.

After the first 128 cycles, it occurs every 75 cycles checking the requirement to load the next column of samples for each of the 3 colors, with the samples appropriate loads during the subsequent 75 cycles. But nevertheless, the initial setting of whether it should be charged during the first set of 75 cycles is always 1 for each color. This allows that the 6th final column of each color within the Memory Intermediate 1 be filled.

At the end of each 75 cycle period, the KernelXCarryOut flag of each color plane of the Generator Resampler Function Address 112 is checked for Determine if the next column of samples should be read. From similarly, an AdvanceLine pulse restarts the process on the Next line if KernelYCarryOut has been established.

Since each "reading" includes effectively 6 or 13 readings to fill a column in the Report Intermediate 1, we maintain an initial position in order to move forward until the next "reading". We also maintain a value of coordinate to allow coordinate generation out-of-bounds that allow Marginal pixel replication, constant color and reset cyclic image.

We consider the active image 180 as being within particular limits, with certain actions that they must be taken when the coordinates are outside the area active The coordinates can be either before the image, or either being inside the image, or after the image, both in terms of lines and pixels. This is shown in the Figure 67, although the space outside the active zone has been exaggerated for reasons of clarity.

Note that because we use (0, 0) As the start of coordinate generation, MaxPixel and MaxLine are also pixel and line counts. However, since the address generation is executed from the realizations of the AdvanceLine function and pulses from the MJI 15, these limits External are not necessary. Address generation for a line simply continues until the AdvanceLine pulse is received, and may include marginal replication, constant colors to outside of limits, or cyclic image pixel reset.

If we have an address, Adr, of the current sample, and we want to move to the next sample, now either on the next line or on the same line, the Sample coordinates will change as expected, but the shape in which direction changes depends on whether we are restarting cyclically the active image, and should produce pixel replication marginal when needed.

When there is no cyclic restart of the image (i.e. all print formats except Passport 31), we carry out the actions of Table 38 as we proceed along the line or pixel To rotate an image at 90 degrees, CPU 10 simply change the values \ DeltaLine and \ Deltapixel.

Observing Table 38, the only time ADR change is using \ Deltapixel when PixelSense is 0, and using \ DeltaLine when LineSense is 0. Following these simple rules, Adr will be valid for marginal pixel replication. Of course, if a constant color is desired outside the coordinates of the limits, you can choose that value instead of the value stored in the appropriate address.

To allow cyclic restart, we compare simply the previous direction (-, 0, +) for Line and Pixel with the new sense When the meaning is "-", we use the advance as described in Table 38, but when the ordinate falls out of limits (that is, moving from 0 to +), we update the Adr

TABLE 38 Actions to be carried out when Advancing in Pixel or Line

59

with a new value that is not based on a delta. Assuming we maintain the initial address for the current line, so that we can advance to the beginning of the next line once the current line has been generated, we can do it next:

?

        If the change is in Pixel, and the direction of the pixel changes from 0 to + (indicating that we have gone beyond the edge of the image), we substitute Adr for LineStartAdr and replace Pixel with ActiveStartPixel. Line is It keeps the same.

?

        If the change is in Line, and the direction changes from 0 to + (which indicates that we have gone beyond the edge of the image), we subtract Adr's DeltaColumn, and we replace Line with ActiveStartLine. Pixel stays the same. Deltacolumn is the direction deviation to generate the addresses of (Pixel, ActiveStartLine) from (Pixel, ActiveEndLine-1).

The logic to load the number of samples of the set (either 6 or 13, depending on the color), is shown in the Pseudo-code that follows.

       \ vskip1.000000 \ baselineskip
    

60

       \ vskip1.000000 \ baselineskip
    

The establishment of such variables as FirstSampleLine, FirstSamplePixel and FirstSampleAdr is performed in the address generator section that corresponds to perform banners from the Function Address Generator, as well as also AdvanceLine pulses from the MJI. The logic for this part of address generation is shown in the following pseudo-code:

61

62

       \ newpage
    

10.2.11.3 Registration Summary

There are a number of records that must be determined before printing an image. These are summarized here in Table 39. To rotate an image in 90 degrees, DeltaLine and DeltaPixel values are simply exchanged, and a new DeltaColumn value is provided.

TABLE 39 Records Required to Be Determined using the Caller before Printing

63

64

11 References

[1] Silverbrook Research, 1998 , Authentication of Consumables

[2] Silverbrook Research, 1998 , Authentication Chip.

Although the invention has been described with Reference to specific examples, subject matter experts You will appreciate that it can materialize in many other ways. The numbered paragraphs that follow provide the addresses with a better indication of the scope of the invention, although others novel features and combination of features they will also appear clearly from the description that follow.

Claims (23)

1. A procedure to provide a image for printing at a point resolution default bi-level that corresponds to a default continuous tone resolution, including Procedure the stages of: receive a first set of indicative data of the image, being the first data set in Bayer format planarized comprising different planes (45, 46, 47) of color, having at least one (46) of the color planes a resolution different from that of the other plane or the other planes (45, 47) of color; perform reconstruction and resampling (64) of image of each color plane of the first data set, to create a second set of continuous tone resolution data default that is greater than the resolution of each of the color planes of the first data set; converting (67) the second data set into a third data set of the predetermined bi-level point resolution that is greater than the predetermined continuous tone resolution of the second data set,
Y make the third set of data be find available for a printer (69) at the resolution of default bi-level point. 2. A procedure in accordance with the claim 1, wherein the first data set is in RGB red, green and blue format, and the printer acts on response to a cyan, magenta and yellow CMY format, including the procedure the stage of converting (66) the second set of data from an RGB format to a CMY format. 3. A procedure in accordance with the claim 1, which includes the step of intensifying (65) the Second set of data. 4. A procedure in accordance with the claim 1, wherein the first data set is obtained from a sensor device, and the procedure includes the stage of compensating the first set of data with respect to the non linearities in the sensor device. 5. A procedure in accordance with the claim 4, wherein the compensation stage includes convert the first data set from a plurality of "x" bit samples to a plurality of "y" samples of bits, being x> y. 6. A procedure in accordance with the claim 5, wherein x = 10 and y = 8. 7. A procedure in accordance with the claim 1, which includes the additional steps of: determine with respect to the first data set, m% of darker pixels and n% of pixels with more brightness; adjust the first data set to match m% of darker pixels, and adjust the first data set to match The n% of pixels with more brightness. 8. A procedure in accordance with the claim 1, which includes the additional step of adjusting the first set of data to provide a balance (62) of the default white. 9. A procedure in accordance with the claim 1, which includes the additional step of adjusting the first data set to provide an expansion (63) of default range 10. A procedure in accordance with the claim 8 or claim 9, wherein the resolution of each color plane of the first data set is incremented while maintaining the same spatial resolution. 11. A procedure in accordance with the claim 1, wherein the first data set is adjusted selectively to provide the image with an orientation default rotational. 12. An apparatus for providing an image for printing at a bi-level point resolution default that corresponds to a continuous tone resolution default, including the device: means of entry to receive a first set of data indicative of the image, the first being data set in planarized Bayer format comprising different planes (45, 46, 47) of color, with at least one (46) of the color planes of a different resolution than the other plane or the other planes (46, 47) of color,

         \ newpage
      

sampling means for reconstruction and resampling (64) image of the first data set, to create a second set of data with a continuous tone resolution default that is greater than the resolution of each of the color planes of the first data set, processing means to convert (67) the second set of data in a third set of data with the default bi-level point resolution which is greater than the default continuous tone resolution of the second data set, and exit means to make the third data set is available (69) so that a Printer print at bi-level dot resolution default 13. An apparatus according to claim 12, in which the first data set is in RGB format of red, green and blue, and the printer acts in response to a format CMY of cyan, magenta and yellow, turning (66) the means of processing the second set of data from an RGB format to a CMY format 14. An apparatus according to claim 12, which also includes an intensification filter for step up (65) the second set of data. 15. An apparatus according to claim 12, in which the first set of data is obtained from a sensor device, and the input means compensates for the first data set for nonlinearities in the device sensor. 16. An apparatus according to claim 15, in which compensation with respect to non-linearities includes converting the first data set from a plurality from "x" bit samples to a plurality of "y" samples of bits, being x> y. 17. An apparatus according to claim 16, in which x = 10 and y = 8. 18. An apparatus according to claim 12, in which the means of entry: determine in relation to the first set of data, the darkest m% of pixels and the n% more pixels bright adjust the first data set to match m% of darker pixels, and adjust the first data set to match The n% of brightest pixels. 19. An apparatus according to claim 12, in which the input means adjust the first set of data to provide a white balance (62) predetermined. 20. An apparatus according to claim 12, in which the input means adjust the first set of data to provide a range expansion (63) default 21. An apparatus according to claim 19 or claim 20, wherein the input means increase the resolution of each color plane of the first set of data while maintaining the same spatial resolution. 22. An apparatus according to claim 12, in which the input means selectively adjust the first set of data to provide the image with a default rotational orientation. 23. A camera that includes: a CCD set to provide an image Bayer; a printer to selectively provide a printed image, and an apparatus according to claim 12, to receive the Bayer image and provide the printer with third set of data so that the image is produced printed.